Ready-to-Use Activity Recognition for Smartphones

Size: px
Start display at page:

Download "Ready-to-Use Activity Recognition for Smartphones"

Transcription

1 Ready-to-Use Activity Recognition for Smartphones Pekka Siirtola, Juha Röning Computer Science and Engineering Department P.O. BOX 4500, FI-90014, University of Oulu, Oulu, Finland Abstract In this study, every day activities are recognized from data collected using smartphones accelerometer sensors. Offline experiments are made to show that the presented method is user- and body position-independent. In addition, it is shown that the features used in the classification are not dependent on the calibration of the phone. The recognition models trained using the offline data are also tested online. A mobile application running these models is built for two operating systems: Symbianˆ3 and Android. Real-time experiments using these applications are made to show that the presented method can be implemented to any operating system and hardware variations do not affect recognition results. Quite high recognition accuracies are obtained, in the offline study, the average recognition rate is almost 99% and, also, in the online study, the average recognition accuracy is over 90%. Index Terms Accelerometer, mobile phones, activity recognition, machine learning I. PROBLEM STATEMENT AND RELATED WORK Nowadays, almost every one has a smartphone and people have them almost all the time with them. Therefore, smartphones are excellent devices to monitor every day activities of people. In this study, every day activities are recognized from acceleration data collected using a mobile phone s sensors. In our previous study (see [1]), we showed that activities can be recognized accurately, in real-time and user-independently using mobile phones. Limitation of that study was that it was not body position-independent, the phone needed to be at trousers front pocket in order to classify movements correctly and reliable. Obviously, the recognition would be much more practical if it would not be dependent on the phone position, which is studied is this article. However, when positionindependent recognition was started to study for this article, challenges were faced. The offline recognition rates were quite high and promising, but when a mobile application was built d based on the trained models, the classification produced weird results and the accuracy was very low. In addition, when the mobile application was tested on multiple devices, different results were obtained with different devices. It was noted that this was caused by differencies in hardware and calibration. Although the phones were of the same model, still there was some differences in accelerometers, and they did not give the same output. This was not an issue in our previous study because the classification process was easier. However, position-independent recognition requires more accurate features and models and therefore hardware variations needed to be considered. Other research have noticed the hardware variation problem as well, in [2] accelerometers were calibrated in order to compensate hardware variations. The study offers two ways to calibrate sensors, a user driven and automatic method. In the user driven calibration method, the user is told to hold the phone on a certain position, and based on this information, variations in sensor values caused by different accelerometer alignments can be calculated. The automatic calibration method tries to find samples where the phone is stationary, and calibration is done based on these samples. According to the study, it takes approximately one or two days to collect samples which can be used to calibrate sensors. Moreover, in the article, five different activities are recognized from the accelerometer data collected using mobile phones. The recognition method presented in the study is body position independent, so the method recognizes activities regardless of the mobile phone s position. Even then, the recognition rates are quite high, 95% of the activities can be recognized correctly in the offline experiments. There are also several other studies ([3], [4], [5], [6], [7]) where the data for activity recognition is collected using a mobile phone but the activity recognition algorithms are not implemented on the phone, and the classification is not done in real-time on a mobile phone as in our study. An activity recognition system running on a smartphone is presented in [8]. The presented system can be trained on the device and does the classification in real-time on the device. The recognition is based on features calculated using geometric template matching. Support vector machine is used as a classifier. Unfortunately, the study does not include recognition rates: thus, the evaluation of the system is difficult. However, the smartphone application is available from Google Play. The system described in [9] can also be found from Google Play. It seems to recognize activities with accurately, but all the features used in classification are not phone orientation independent. This paper present an activity recognition method that takes into consideration hardware variations. However, it does not require calibration, user inputs are not needed and a mobile application based on the method is ready-to-use straight after installation. Instead, we focus on considering hardware variations already in the feature extraction phase. In addition, it is shown using offline tests that these features can be used to rec-

2 ognize human activities body position- and user-independently. What is more, the real-time experiments show that the presented method can be implemented to every smartphone where the maximum sampling rate of the accelerometers is at least 40Hz. The paper is organized as follows: Section II presents the used data sets. Section III introduces the methods and Section IV evaluates the accuracy of the proposed method in offline and online scenarios. Finally, the conclusions are in Section V. II. DATA COLLECTION The data were collected using a Nokia N8 smartphone [10] running Symbianˆ3 operating system. N8 includes a wide range of sensors: tri-axis accelerometer and magnetometer, two cameras (12 MP and 0.3 MP), GPS, proximity sensor, compass, microphones and ambient light sensor. Still, only the triaxis accelerometer was used in this study to detect activities. Different phones have different hardware, and therefore, the maximum sampling rate at smartphones can differ. Thus, in order to achieve mobile phone independent results, the highest possible sampling rate was not used. Although accelerometers were running at full speed during the collection of the training data, all the samples were not saved nevertheless, and therefore, not used in activity recognition process. Instead, the latest value from the accelerometer was called at every 25 milliseconds and stored to a file. Therefore, the used sampling frequency was 40Hz, which is much less than the maximum sampling frequency of most phones. The used method enables the same sampling frequency to any smartphone, where the maximum frecuency of the accelerometer is at least 40Hz, making recognition less phone model dependent. The classification models used in this study were trained based on the activity data collected from seven healthy subjects. The subjects were carrying five phones at the same time. They were located at trousers front pocket, jacket s pocket, at backpack, at brachium and one at the ear. The participants performed five different activities: walking, running, cycling, driving a car, and sitting/standing. The reason for selecting these activities for the study is that normal everyday life consists mainly of these five activities. Howver, there is no data from each activity from each body position, see I. For instance, subjects were not allowed to cycle while holding a phone at the ear because of safety issues. Moreover, data were collected when a phone was laying on the table. Therefore, six activities were recognized. The total amount of the data collected was about fifteen hours. When the selected activities are studied in more detail, it can be noted that walking and running are different from the other three because everyone has a personal walking and running style. Other activities are not personal, for instance, while cycling, the movement trajectory is predefined. Therefore, the models to recognize walking and running are most challenging to train. The training data were collected by seven subjects whose age varied from 24 to 34 years (average 29 years) and height TABLE I ACCELEROMETER DATA WERE COLLECTED FROM FIVE DIFFERENT BODY LOCATIONS AND FROM SIX DIFFERENT ACTIVITIES. Body position trouser s pocket jacket s pocket backpack ear brachium table Activities walking, running, cycling, sitting/standing, driving a car walking, running, cycling, sitting/standing walking, running, cycling, sitting/standing walking walking, running, cycling, sitting/standing on the table from 1.65 to 1.85 meter (average 1.78 meter). They performed most of the activities outside the laboratory. Subjects walked inside and outside, mainly on flat surface, but also in a staircase. Streets where subjects walked, run, drove a car, and cycled were normal Finnish tarmac roads, and the route and speed were determined by subjects themselves. In addition, the phone orientation was determined by subjects. The roads used for collecting driving a car data included motorways, as well as roads at the city center. Partly the same roads were employed in offline and real-time tests. Sitting consists mostly of office working. In addition, the real-time classification on the device was tested to show that the presented method is not dependent on hardware and operating system. For this experiment, the activity recognition application was implemented to Symbianˆ3 and Android -operating systems. Symbianˆ3 tests were made using Nokia N8 smartphone and Android tests using Samsung Galaxy Mini smartphone running Android operating system. Galaxy Mini is a low budget smartphone having triaxis accelerometer, proximity sensor, compass and 3.15MP camera. It uses 600MHz ARMv6 processor. A. Aim of the study III. METHODS The activity recognition system presented in this study is ready-to-use. This means that the user does not need to set any parameters or do any other actions before he/she starts to use it. Therefore, the system presented in this study is user-independent. In addition, the study uses calibration and orientation independent features, and therefore, the system can be used regardless the hardware differences. Moreover, unlike in our previous study, the user does not need to carry a phone on trousers pocket. Because training data was collected from five different body locations, the presented method supports five different body locations, and therefore, it can be said that it is body-position independent. What is more, using the presented method, an activity recognition system can be implemented to any smartphone, where the frequency of the accelerometer is atleast 40Hz, making the system mobile phone independent. B. Activity recognition Activity recognition process consists of four stages: preprocessing, feature extraction, model training, and classification.

3 In this study, the aim is to do the whole classification process on the device in real-time. Therefore, the recognition process needs to be light so that it does not require a lot of processing power and does not use a lot of battery. 1) Sliding window: The online activity recognition was done using a sliding window technique. The signals from the sensors were divided into equal-sized smaller sequences, also called windows. From these windows, features were extracted and finally the classification of the sequences was done using models trained based on these features. In this study, the windows were of the length of 300 observations, which is 7.5 seconds, because the sampling frequency was 40Hz. In offline recognition, the slide between two sequential windows was 75 observations, while in online recognition, the slide was set to 150 observations to load the processor less. Moreover, to reduce the number of misclassified windows, the final classification was done based on the majority voting of the classification results of three adjacent windows. Therefore, when an activity changes, a new activity can be detected when two adjacent windows are classified as a new activity. For instance, if the slide is 150 observations, a new activity can be detected after 450 observations, which is around eleven seconds if the sampling rate is 40Hz. 2) Feature extraction: In order to achieve reliable phone orientation independent results, the effect of gravitation needs to be elimanated from the sensor values. There are two ways to do that: (1) by recognizing the orientation of the smartphone, or (2) by eliminating the orientation information. On the other hand, the orientation is impossible to recognize using only accelerometers. Therefore, the effect of orientation had to be eliminated. In the preprocessing stage, all three acceleration channels were combined as one using square summing. This way orientation independent magnitude acceleration was obtained. Moreover, the orientation of the phone has some limitations, for example the screen or the back of the smartphone is always against the user s leg when the phone is in the trousers pocket. Therefore, it was tested if features extracted from a signal where two out of three acceleration channels were square summed would improve the classification accuracy. A 3D accelerometer is a sensor consisting of three accelerometers that are approximately perpendicular to each other. However, in reality, in each phone these sensors are aligned a bit differently. Therefore, the output data of the sensors is dependent on the used phone making also the classification result dependent on the phone. Figure 1 shows a magnitude signal from three different N8 smartphones when the phone is still on the table. Average values vary from 9.1 to 10.0 m/s 2 and this variation causes misclassifications. In order to achieve the highest possible detection rates, the methods presented in the study take into considedation these hardware variations. However, unlike previous studies ([2]), it does not require calibration. Instead, we focus on considering hardware variations in the feature extraction phase. Therefore, user inputs are not needed and the mobile application based on the method is ready-to-use straight after installation. Because of the different calibration, accelerometer values from different phones are at a different level. Therefore, the shapes of the signals are the same but absolute values differ by some constant. This difference is normally fixed by using automated or user-driven calibration. However, because the difference in the signal level is the only major difference caused by different calibration, in this study, differences are eliminated simply by subtracting the mean of the window s values from each value of the window. This way, calibration differences can be eliminated already in the feature extraction phase and calibration is not needed. 19 features were extracted from magnitude signal and from the signals combining two out of three acceleration channels after calibration differences were eliminated, so together 76 features were extracted. These features were standard deviation, minimum, maximum. In addition, instead of extracting percentiles, the remainder between percentiles (10, 25, 75, and 90) and median were calculated. Moreover, the sum of values above or below percentile (10, 25, 75, and 90), square sum of values above of below percentile (10, 25, 75, and 90), before summing and number of crossings above of below percentile (10, 25, 75, and 90) were extracted and used as features. Note that the mean was not one of the features because it is zero for every window caused by eliminating calibration differences. 3) Classification: The classification result was obtained using the decision tree presented in Figure 2, which classifies activities using a three stage procedure. In the first classification stage, a model is trained to decide if the studied subject is currently active (walking, running or cycling) or inactive (driving a car, sitting, standing or a phone on the table). After this, one or two more classifications are needed to perform in order to obtain the exact activity. Therefore, five binary classifiers are needed to train. These models were implemented to a smartphone and also used in online tests. Based on the results of our previous study, QDA (quadratic discriminant analysis) was decided to use as a classifier. QDA combines features to find quadric surfaces that separate the classes best. The resulting combination may be employed as a quadric classifier [11]. In order to achieve the highest possible recognition rates, the most descriptive features for each model were selected using a sequential forward selection (SFS) method [12]. Moreover, to obtain reliable user-independent results, the training was performed using the leave-one-out method, so that each person s data in turn was used for testing and the rest of the data were employed for model training. IV. EXPERIMENTS The experiments section comprises two parts: offline recognition and online recognition. The purpose of the offline recognition is to show that the presented method is userindependent and, in addition to showing that it is body position-independent. The models used to offline recognition are then implemented to a mobile phone and online experiments are made to show that the presented method is reliable regardless of the operating system and the hardware variations of the mobile phones. Moreover, experiments are made to

4 Fig. 1. Different phones have different calibration. Fig. 2. The decition tree obtained to recognize the type of activity. TABLE III OFFLINE EXPERIMENTS SHOW THAT ACTIVITIES ARE RECOGNIZED WITH HIGH ACCURACY FROM EVERY BODY LOCATION. Body position Recognition rate trouser s pocket 99.5 % jacket s pocket 97.3 % backpack 97.8 % ear 96.7 % brachium 98.9% table 98.7 % show that the method presented is light enough to be used for real-time recognition. The models were trained using the QDA classifier and they were based on the data set presented in Section II. Online experimets are performed by a person carrying six different smartphones. A. Offline recognition The offline results are based on the data presented in Section II. Model training was performed using the leave-one-out method, so that each person s data in turn was used for testing and the data of six persons were employed for model training. 1) Results: User- and body position -independent results are presented in Table II. To compare how activities are recognized from different body positions, the recognition accuracy from each body position was obtained and the results are presented in Table III. 2) Discussion: Offline experimets show that using the presented method activities are recognized with a high accuracy user-independently, see Table II. In addition, all the activities are detected almost perfectly from each body-position. Moreover, the average recognition rate (98.9%) is even higher than in our previous study despite the fact that, in this study, there is one activity more and activities are recognized body position-independently. However, the results from these two studies are not fully comparable because of different data sets. In additon, hardware variations between phones is not an issue, this can be seen from the results because the training data is collected using five different smartphones and activities from each phone are detected reliably. To show that hardware variations are not an issue in real-time on device recognition either, activity recognition applications for Symbianˆ3- and Android- operating systems based on the models used in offline tests were also coded. B. Online recognition An activity recognition application for Symbianˆ3 devices was built using Qt [13] programming language. In addition, the Android-version of the activity recognition application was built using Java programming language and tested using Samsung Galaxy Mini smartphone. The purpose of the online recognition experiment is to show that the presented method is not dependent on hardware and operating system. 1) Results: Online recognition on the device was tested by a person carrying several devices: five Nokia N8 smartphones and Samsung Galaxy Mini. He carried these phones at each of the five body locations. The recognition rates of different activities and recognition rates of activities from different body positions using Nokia N8 are presented in Table IV and the ones using Samsung Galaxy Mini in Table V. 2) Discussion: Online recognition accuracies are high, however, not as high as offline recognition results. This is understandable because it is impossible to reckon with all the situations of every day life. Now, for instance, the recognition rates for driving a car in the online case were not as high as in the model training phase, bacause there were a lot of red traffic lights. These stops were classified as sitting instead of driving a car, and thus, making the detection rate for drivingactivity the lowest. In addition, while cycling, people tend to just taxi at times, so they do not pedal constantly. Therefore, it is a matter of an opinion if one concludes these events as cycling or not. This sort of situation was tested and it seems that these events are classified as driving a car. This is logical because both events cause quite similar vibrations due to the roughness of the road. Real-time recognition was tested using Android- and Symbianˆ3- phones. According to the results presented in

5 TABLE II OFFLINE ACTIVITY RECOGNITION RESULTS USING QDA. True class / predicted sitting/standing walking cycling running on table driving class sitting/standing 92.0% 0.1 % 0.0 % 0.0 % 3.0 % 4.9 % walking 0.0 % 98.0% 1.7% 0.3 % 0.0 % 0.0 % cycling 0.0 % 1.3 % 98.7% 0.0 % 0.0 % 0.0 % running 0.0 % 0.0 % 0.0 % 100.0% 0.0 % 0.0 % on table 1.3 % 0.0 % 0.0 % 0.0 % 98.7% 0.0 % driving 0.0 % 0.0 % 0.0 % 0.0 % 0.0 % 100.0% TABLE IV REAL-TIME ON DEVICE RECOGNITION RATES OF ACTIVITIES AND THE RECOGNITION RATES OF ACTIVITIES FROM DIFFERENT BODY POSITIONS USING SYMBIANˆ3 SMARTPHONE. Activity Recognition rate Body position Recognition rate walking 94.5 % trousers 94.2 % cycling 93.5 % jacket 94.9 % running % backpack 92.9 % sitting/standing 90.8 % ear % driving 84.5% brachium 93.1 % on the table 99.2% table 99.2 % TABLE V REAL-TIME ON DEVICE RECOGNITION RATES OF DIFFERENT ACTIVITIES AND THE RECOGNITION RATES OF ACTIVITIES FROM DIFFERENT BODY POSITIONS USING ANDROID SMARTPHONE. Activity Recognition rate Body position Recognition rate walking 94.4 % trousers 90.8 % cycling 91.8 % jacket 90.2 % running 95.4 % backpack 96.3 % sitting/standing 89.0 % ear 100.0% driving 85.1% brachium 90.5 % on the table 93.2% table 93.2 % Tables IV and V, there is not much difference in the recognition rates between operating systems. Neither when the rates between activities are compared, nor when the rates between body positions are compared. In fact, in both cases, the difference is not statistically significant according to the paired t-test. However, sitting/standing -activity was an exception. Both operating systems recognized it with approximately same accuracy but misclassification were classified to different classes. Symbian misclassified it to on table -activity while Android misclassified it as driving. It should be further study what caused this difference. However, the results are not fully comparable because not all phones were carried at the same time. Still, it seems that different hardware and operating system do not affect heavily the detection rates. However, in can be said that the presented activity recognition system is hardware and operating system independent. It works with every smartphone where the maximum sampling rate of the acceleration sensors is higher than 40Hz. It is light, too, Symbianˆ3 -version of the application uses around 15% of CPU capacity of the Nokia N8; thus, other applications can be run alongside. V. CONCLUSIONS This study presented an ready-to-use activity recognition system, it does not require any calibration or user inputs. Offline experiments were made using a data set collected from seven subjects. The experiment showed that the presented method is user-independent, activities from every subject were recognized with a high accuracy. In addition, this experiment showed that activities can be classified reliably regardless of the phone position and orientation. What is more, the online experiments done using six smartphones running two different operating systems, and because there is no significant difference between detection accuracies between operating systems, the recognition system is not dependent on the phone hardware or operating system because of the calibration independent features. Also, the different maximum sampling rates of different phones are not an issue, because in this study, the maximum sampling rate was not used. Although the accelerometers were running at full speed, a new value to be used in activity detection was nevertheless called at every 25 millisecond. Therefore, the used frequency was 40Hz, which is much less than the maximum frequency of most of the smartphones. Thus, the presented method can be used with every smartphone and it is not dependent on the phone model. While in the offline scanerio activities are classified almost perfectly, the real-time on device recognition results are not quite as high. However, also the online recognition rates are high, over 90%, with both tested operating systems. The difference between recognition rates of online and offline scenarios

6 is most likely caused by real-life situations from which there is no training data. Also other researchers have noted that reallife recognition rates are often lower than offline accuracies. In [14], it is studied how the recognition rate decreases when activity recognition models trained using the supervised data are tested with unsupervised data. In the study, the model was trained to recognize nine different sports/everyday activities and when it was tested using supervised data, the recognition rate was 91%. When the same model was tested using unsupervised data, including four out of nine trained activities, the average recognition rate was only 64%. When these recognition rates are compared to the results of this study, it can be noted that online recognition works quite well. Moreover, the detection accuracies can be further improved by building a behavior recognition system based on the activity classification results. Although the purpose of the real-time on device experiment was to show that activity recognition works regardless of the hardware and operating system differences, in future studies, it should be tested by a larger amount of test subjects. Now only one test person was used. In addition, it would be beneficial to collect more training data as well. This could reduce the differences between online and offline recognition rates. Moreover,null-data recognition is not included in this study, either, and therefore, such activities cause incorrect classifications. Thus, to improve the accuracy of the application, the nullactivity recognition should be included. In this study, everything except model training is done on the device. Another option would be to send the sensor data to the server, perform the classification process there and, finally, send the results to a smartphone using 3G. In this case, calculation capacity would not be an issue, but on the other hand, privacy issues should be handled. In addition, data transfer is not free and can cause exceptionally high costs, especially abroad. [4] D. Peebles, H. Lu, N. D. Lane, T. Choudhury, and A. T. Campbell, Community-guided learning: Exploiting mobile sensor users to model human behavior, in Proceedings of the Twenty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2010, Atlanta, Georgia, USA, July 11-15, [5] J. Ryder, B. Longstaff, S. Reddy, and D. Estrin, Ambulation: A tool for monitoring mobility patterns over time using mobile phones, in Computational Science and Engineering, CSE 09. International Conference on, vol. 4, aug. 2009, pp [6] L. Sun, D. Zhang, B. Li, B. Guo, and S. Li, Activity recognition on an accelerometer embedded mobile phone with varying positions and orientations, in Ubiquitous Intelligence and Computing, ser. Lecture Notes in Computer Science. Springer Berlin / Heidelberg, 2010, vol. 6406, pp [7] S. Wang, C. Chen, and J. Ma, Accelerometer based transportation mode recognition on mobile phones, in Wearable Computing Systems (APWCS), 2010 Asia-Pacific Conference on, april 2010, pp [8] J. Frank, S. Mannor, and D. Precup, Activity recognition with mobile phones, in Machine Learning and Knowledge Discovery in Databases, ser. Lecture Notes in Computer Science. Springer Berlin / Heidelberg, 2011, vol. 6913, pp [9] J. R. Kwapisz, G. M. Weiss, and S. A. Moore, Activity recognition using cell phone accelerometers, SIGKDD Explor. Newsl., vol. 12, pp , March [10] Nokia N8, [11] D. J. Hand, H. Mannila, and P. Smyth, Principles of data mining. Cambridge, MA, USA: MIT Press, [12] E. Haapalainen, P. Laurinen, H. Junno, L. Tuovinen, and J. Röning, Feature selection for identification of spot welding processes, Proceedings of the 3rd International Conference on Informatics in Control, Automation and Robotics, pp , [13] Qt, [14] M. Ermes, J. Pärkkä, J. Mäntyjärvi, and I. Korhonen, Detection of daily activities and sports with wearable sensors in controlled and uncontrolled conditions, Information Technology in Biomedicine, IEEE Transactions on, vol. 12, no. 1, pp , Jan ACKNOWLEDGMENT This work was done as a part of MOPO study (Clinical- Trials.gov Identifier: NCT ). The authors would like to thank Infotech Oulu and the Finnish Funding Agency for Technology and Innovation for funding this work. Special thanks to Tero Vallius who helped with the coding work and to Johanna Pyy for collecting the data. REFERENCES [1] P. Siirtola and J. Röning, Recognizing human activities userindependently on smartphones based on accelerometer data, International Journal of Interactive Multimedia and Artificial Intelligence, vol. 1, no. 5, pp , 06/ [2] H. Lu, J. Yang, Z. Liu, N. D. Lane, T. Choudhury, and A. T. Campbell, The Jigsaw continuous sensing engine for mobile phone applications, in Proceedings of the 8th ACM Conference on Embedded Networked Sensor Systems, ser. SenSys 10, 2010, pp [3] T. Brezmes, J.-L. Gorricho, and J. Cotrina, Activity recognition from accelerometer data on a mobile phone, in Distributed Computing, Artificial Intelligence, Bioinformatics, Soft Computing, and Ambient Assisted Living, ser. Lecture Notes in Computer Science. Springer Berlin / Heidelberg, 2009, vol. 5518, pp

Activity recognition using a wrist-worn inertial measurement unit: a case study for industrial assembly lines

Activity recognition using a wrist-worn inertial measurement unit: a case study for industrial assembly lines Activity recognition using a wrist-worn inertial measurement unit: a case study for industrial assembly lines Heli Koskimäki, Ville Huikari, Pekka Siirtola, Perttu Laurinen, Juha Röning Intelligent Systems

More information

Activity Recognition Using Cell Phone Accelerometers

Activity Recognition Using Cell Phone Accelerometers Activity Recognition Using Cell Phone Accelerometers Jennifer Kwapisz, Gary Weiss, Samuel Moore Department of Computer & Info. Science Fordham University 1 We are Interested in WISDM WISDM: WIreless Sensor

More information

A novel approach to classify human-motion in smart phone using 2d-projection method

A novel approach to classify human-motion in smart phone using 2d-projection method A novel approach to classify human-motion in smart phone using 2d-projection method 1 Yi Suk Kwon, 1 Yeon Sik Noh, 1 Ja Woong Yoon, 1 Sung Bin Park, 1 Hyung Ro Yoon 1 Department of Biomedical Engineering

More information

Measuring the Capability of Smartphones for Executing Context Algorithms

Measuring the Capability of Smartphones for Executing Context Algorithms Maximilian Eibl, Martin Gaedke. (Hrsg.): INFORMATIK 2017, Lecture Lecture Notes in Notes Informatics in Informatics (LNI), Gesellschaft (LNI), Gesellschaft für Informatik, für Informatik, Bonn 2017 Bonn

More information

A SIMPLE HIERARCHICAL ACTIVITY RECOGNITION SYSTEM USING A GRAVITY SENSOR AND ACCELEROMETER ON A SMARTPHONE

A SIMPLE HIERARCHICAL ACTIVITY RECOGNITION SYSTEM USING A GRAVITY SENSOR AND ACCELEROMETER ON A SMARTPHONE International Journal of Technology (2016) 5: 831-839 ISSN 2086-9614 IJTech 2016 A SIMPLE HIERARCHICAL ACTIVITY RECOGNITION SYSTEM USING A GRAVITY SENSOR AND ACCELEROMETER ON A SMARTPHONE Alvin Prayuda

More information

Video Inter-frame Forgery Identification Based on Optical Flow Consistency

Video Inter-frame Forgery Identification Based on Optical Flow Consistency Sensors & Transducers 24 by IFSA Publishing, S. L. http://www.sensorsportal.com Video Inter-frame Forgery Identification Based on Optical Flow Consistency Qi Wang, Zhaohong Li, Zhenzhen Zhang, Qinglong

More information

Title. Author(s)Takeuchi, Shin'ichi; Tamura, Satoshi; Hayamizu, Sato. Issue Date Doc URL. Type. Note. File Information

Title. Author(s)Takeuchi, Shin'ichi; Tamura, Satoshi; Hayamizu, Sato. Issue Date Doc URL. Type. Note. File Information Title Human Action Recognition Using Acceleration Informat Author(s)Takeuchi, Shin'ichi; Tamura, Satoshi; Hayamizu, Sato Proceedings : APSIPA ASC 2009 : Asia-Pacific Signal Citationand Conference: 829-832

More information

RECOGNITION OF DRIVING MANEUVERS BASED ACCELEROMETER SENSOR

RECOGNITION OF DRIVING MANEUVERS BASED ACCELEROMETER SENSOR International Journal of Civil Engineering and Technology (IJCIET) Volume 9, Issue 11, November 2018, pp. 1542 1547, Article ID: IJCIET_09_11_149 Available online at http://www.iaeme.com/ijciet/issues.asp?jtype=ijciet&vtype=9&itype=11

More information

Tracking driver actions and guiding phone usage for safer driving. Hongyu Li Jan 25, 2018

Tracking driver actions and guiding phone usage for safer driving. Hongyu Li Jan 25, 2018 Tracking driver actions and guiding phone usage for safer driving Hongyu Li Jan 25, 2018 1 Smart devices risks and opportunities Phone in use 14% Other distractions 86% Distraction-Affected Fatalities

More information

Face Recognition Using Vector Quantization Histogram and Support Vector Machine Classifier Rong-sheng LI, Fei-fei LEE *, Yan YAN and Qiu CHEN

Face Recognition Using Vector Quantization Histogram and Support Vector Machine Classifier Rong-sheng LI, Fei-fei LEE *, Yan YAN and Qiu CHEN 2016 International Conference on Artificial Intelligence: Techniques and Applications (AITA 2016) ISBN: 978-1-60595-389-2 Face Recognition Using Vector Quantization Histogram and Support Vector Machine

More information

EMBEDDED SYSTEMS AND MOBILE SYSTEMS

EMBEDDED SYSTEMS AND MOBILE SYSTEMS EMBEDDED SYSTEMS AND MOBILE SYSTEMS Embedded systems Sensors, actuators, devices 8-bit uc, sensors, actuators Mobile systems: Portable PC (powerfull, WiFi connected, heavy, cumbersome, about 4h of battery

More information

Exploring unconstrained mobile sensor based human activity recognition

Exploring unconstrained mobile sensor based human activity recognition Exploring unconstrained mobile sensor based human activity recognition ABSTRACT Luis Gerardo Mojica de la Vega luis.mojica@utdallas.edu Arvind Balasubramanian arvind@utdallas.edu Human activity recognition

More information

mfingerprint: Privacy-Preserving User Modeling with Multimodal Mobile Device Footprints

mfingerprint: Privacy-Preserving User Modeling with Multimodal Mobile Device Footprints mfingerprint: Privacy-Preserving User Modeling with Multimodal Mobile Device Footprints Haipeng Zhang 1, Zhixian Yan 2, Jun Yang 2, Emmanuel Munguia Tapia 2, and David J. Crandall 1 1 School of Informatics

More information

Mobile and Ubiquitous Computing: Mobile Sensing

Mobile and Ubiquitous Computing: Mobile Sensing Mobile and Ubiquitous Computing: Mobile Sensing Master studies, Winter 2015/2016 Dr Veljko Pejović Veljko.Pejovic@fri.uni-lj.si Based on: Mobile and Ubiquitous Computing Mirco Musolesi, University of Birmingham,

More information

REAL-TIME ROAD SIGNS RECOGNITION USING MOBILE GPU

REAL-TIME ROAD SIGNS RECOGNITION USING MOBILE GPU High-Performance Сomputing REAL-TIME ROAD SIGNS RECOGNITION USING MOBILE GPU P.Y. Yakimov Samara National Research University, Samara, Russia Abstract. This article shows an effective implementation of

More information

A Mobile Platform for Measurements in Dynamic Topology Wireless Networks

A Mobile Platform for Measurements in Dynamic Topology Wireless Networks A Mobile Platform for Measurements in Dynamic Topology Wireless Networks E. Scuderi, R.E. Parrinello, D. Izal, G.P. Perrucci, F.H.P. Fitzek, S. Palazzo, and A. Molinaro Abstract Due to the wide spread

More information

CS 528 Mobile and Ubiquitous Computing Lecture 7b: Smartphone Sensing. Emmanuel Agu

CS 528 Mobile and Ubiquitous Computing Lecture 7b: Smartphone Sensing. Emmanuel Agu CS 528 Mobile and Ubiquitous Computing Lecture 7b: Smartphone Sensing Emmanuel Agu Smartphone Sensors Typical smartphone sensors today accelerometer, compass, GPS, microphone, camera, proximity Use machine

More information

Measuring Height of an Object using Accelerometer and Camera in ios and Android Devices

Measuring Height of an Object using Accelerometer and Camera in ios and Android Devices Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 3, Issue. 6, June 2014, pg.380

More information

Programming-By-Example Gesture Recognition Kevin Gabayan, Steven Lansel December 15, 2006

Programming-By-Example Gesture Recognition Kevin Gabayan, Steven Lansel December 15, 2006 Programming-By-Example Gesture Recognition Kevin Gabayan, Steven Lansel December 15, 6 Abstract Machine learning and hardware improvements to a programming-by-example rapid prototyping system are proposed.

More information

Intelligent Mobile Monitoring and Control Network of Oil Pipeline Based on Internet of Things

Intelligent Mobile Monitoring and Control Network of Oil Pipeline Based on Internet of Things Journal of Information Hiding and Multimedia Signal Processing c 2017 ISSN 2073-4212 Ubiquitous International Volume 8, Number 6, November 2017 Intelligent Mobile Monitoring and Control Network of Oil

More information

Machine Learning for the Quantified Self. Lecture 2 Basic of Sensory Data

Machine Learning for the Quantified Self. Lecture 2 Basic of Sensory Data Machine Learning for the Quantified Self Lecture 2 Basic of Sensory Data Dataset (1) During the course we will use a running example provided by CrowdSignals.io People share their mobile sensors data (smart

More information

A RFID System for Recognizing Activity Using Radio Patterns Shwetha R 1, Sushmitha P 2, Swathi MS 3, Poonam Kumari 4

A RFID System for Recognizing Activity Using Radio Patterns Shwetha R 1, Sushmitha P 2, Swathi MS 3, Poonam Kumari 4 A RFID System for Recognizing Activity Using Radio Patterns Shwetha R 1, Sushmitha P 2, Swathi MS 3, Poonam Kumari 4 1 BE Student, CSE, RRCE, Karnataka, India 2 BE Student, CSE, RRCE, Karnataka, India

More information

Structure of Association Rule Classifiers: a Review

Structure of Association Rule Classifiers: a Review Structure of Association Rule Classifiers: a Review Koen Vanhoof Benoît Depaire Transportation Research Institute (IMOB), University Hasselt 3590 Diepenbeek, Belgium koen.vanhoof@uhasselt.be benoit.depaire@uhasselt.be

More information

Development of a mobile application for manual traffic counts

Development of a mobile application for manual traffic counts Development of a mobile application for manual traffic counts Mohammad Ghanim 1,* and Khalid Khawaja 2 1 Department of Civil and Architectural Engineering, Qatar University, Doha, Qatar 2 Office of Academic

More information

Social Behavior Prediction Through Reality Mining

Social Behavior Prediction Through Reality Mining Social Behavior Prediction Through Reality Mining Charlie Dagli, William Campbell, Clifford Weinstein Human Language Technology Group MIT Lincoln Laboratory This work was sponsored by the DDR&E / RRTO

More information

Face Detection using Hierarchical SVM

Face Detection using Hierarchical SVM Face Detection using Hierarchical SVM ECE 795 Pattern Recognition Christos Kyrkou Fall Semester 2010 1. Introduction Face detection in video is the process of detecting and classifying small images extracted

More information

REAL TIME PUBLIC TRANSPORT INFORMATION SERVICE

REAL TIME PUBLIC TRANSPORT INFORMATION SERVICE Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 4, Issue. 7, July 2015, pg.88

More information

The Google Maps app for iphone and ipad makes navigating your world faster and easier. Voice-guided GPS navigation for driving, biking, and walking

The Google Maps app for iphone and ipad makes navigating your world faster and easier. Voice-guided GPS navigation for driving, biking, and walking Google Maps Earth App For Iphone 4 Voice Navigation Try these quick fixes to improve voice-guided navigation in the Google Maps app. Can't hear voice navigation? Here are a couple tips if you're using

More information

Overview. Background. Intelligence at the Edge. Learning at the Edge: Challenges and Brainstorming. Amazon Alexa Smart Home!

Overview. Background. Intelligence at the Edge. Learning at the Edge: Challenges and Brainstorming. Amazon Alexa Smart Home! Overview Background Intelligence at the Edge Samsung Research Learning at the Edge: Challenges and Brainstorming Amazon Alexa Smart Home! Background Ph.D. at UW CSE RFID, Mobile, Sensors, Data Nokia Research

More information

A Supervised Method for Multi-keyword Web Crawling on Web Forums

A Supervised Method for Multi-keyword Web Crawling on Web Forums Available Online at www.ijcsmc.com International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology IJCSMC, Vol. 3, Issue. 2, February 2014,

More information

Online Pose Classification and Walking Speed Estimation using Handheld Devices

Online Pose Classification and Walking Speed Estimation using Handheld Devices Online Pose Classification and Walking Speed Estimation using Handheld Devices Jun-geun Park MIT CSAIL Joint work with: Ami Patel (MIT EECS), Jonathan Ledlie (Nokia Research), Dorothy Curtis (MIT CSAIL),

More information

Global Journal of Engineering Science and Research Management

Global Journal of Engineering Science and Research Management A REVIEW PAPER ON INERTIAL SENSOR BASED ALPHABET RECOGNITION USING CLASSIFIERS Mrs. Jahidabegum K. Shaikh *1 Prof. N.A. Dawande 2 1* E & TC department, Dr. D.Y.Patil college Of Engineering, Talegaon Ambi

More information

Coin Size Wireless Sensor Interface for Interaction with Remote Displays

Coin Size Wireless Sensor Interface for Interaction with Remote Displays Coin Size Wireless Sensor Interface for Interaction with Remote Displays Atia Ayman, Shin Takahashi, and Jiro Tanaka Department of Computer Science, Graduate school of systems and information engineering,

More information

The Un-normalized Graph p-laplacian based Semi-supervised Learning Method and Speech Recognition Problem

The Un-normalized Graph p-laplacian based Semi-supervised Learning Method and Speech Recognition Problem Int. J. Advance Soft Compu. Appl, Vol. 9, No. 1, March 2017 ISSN 2074-8523 The Un-normalized Graph p-laplacian based Semi-supervised Learning Method and Speech Recognition Problem Loc Tran 1 and Linh Tran

More information

The Impact of Feature Vector Length on Activity Recognition Accuracy on Mobile Phone

The Impact of Feature Vector Length on Activity Recognition Accuracy on Mobile Phone , July 1-3, 2015, London, U.K. The Impact of Feature Vector Length on Activity Recognition Accuracy on Mobile Phone Sulaimon A. Bashir, Daniel C. Doolan, Andrei Petrovski Abstract A key challenge for large

More information

Non-rigid body Object Tracking using Fuzzy Neural System based on Multiple ROIs and Adaptive Motion Frame Method

Non-rigid body Object Tracking using Fuzzy Neural System based on Multiple ROIs and Adaptive Motion Frame Method Proceedings of the 2009 IEEE International Conference on Systems, Man, and Cybernetics San Antonio, TX, USA - October 2009 Non-rigid body Object Tracking using Fuzzy Neural System based on Multiple ROIs

More information

Detecting Anomalous Trajectories and Traffic Services

Detecting Anomalous Trajectories and Traffic Services Detecting Anomalous Trajectories and Traffic Services Mazen Ismael Faculty of Information Technology, BUT Božetěchova 1/2, 66 Brno Mazen.ismael@vut.cz Abstract. Among the traffic studies; the importance

More information

Color-Based Classification of Natural Rock Images Using Classifier Combinations

Color-Based Classification of Natural Rock Images Using Classifier Combinations Color-Based Classification of Natural Rock Images Using Classifier Combinations Leena Lepistö, Iivari Kunttu, and Ari Visa Tampere University of Technology, Institute of Signal Processing, P.O. Box 553,

More information

Detecting Harmful Hand Behaviors with Machine Learning from Wearable Motion Sensor Data

Detecting Harmful Hand Behaviors with Machine Learning from Wearable Motion Sensor Data Detecting Harmful Hand Behaviors with Machine Learning from Wearable Motion Sensor Data Lingfeng Zhang and Philip K. Chan Florida Institute of Technology, Melbourne, FL 32901 lingfeng2013@my.fit.edu, pkc@cs.fit.edu

More information

Keywords Data alignment, Data annotation, Web database, Search Result Record

Keywords Data alignment, Data annotation, Web database, Search Result Record Volume 5, Issue 8, August 2015 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Annotating Web

More information

Traffic Signs Recognition using HP and HOG Descriptors Combined to MLP and SVM Classifiers

Traffic Signs Recognition using HP and HOG Descriptors Combined to MLP and SVM Classifiers Traffic Signs Recognition using HP and HOG Descriptors Combined to MLP and SVM Classifiers A. Salhi, B. Minaoui, M. Fakir, H. Chakib, H. Grimech Faculty of science and Technology Sultan Moulay Slimane

More information

IN computer vision develop mathematical techniques in

IN computer vision develop mathematical techniques in International Journal of Scientific & Engineering Research Volume 4, Issue3, March-2013 1 Object Tracking Based On Tracking-Learning-Detection Rupali S. Chavan, Mr. S.M.Patil Abstract -In this paper; we

More information

Research Fellow, Korea Institute of Civil Engineering and Building Technology, Korea (*corresponding author) 2

Research Fellow, Korea Institute of Civil Engineering and Building Technology, Korea (*corresponding author) 2 Algorithm and Experiment for Vision-Based Recognition of Road Surface Conditions Using Polarization and Wavelet Transform 1 Seung-Ki Ryu *, 2 Taehyeong Kim, 3 Eunjoo Bae, 4 Seung-Rae Lee 1 Research Fellow,

More information

Predicting Physical Activities from Accelerometer Readings in Spherical Coordinate System

Predicting Physical Activities from Accelerometer Readings in Spherical Coordinate System Predicting Physical Activities from Accelerometer Readings in Spherical Coordinate System Kittikawin Lehsan and Jakramate Bootkrajang* Department of Computer Science, Chiang Mai University Chiang Mai,

More information

Fully Automatic Methodology for Human Action Recognition Incorporating Dynamic Information

Fully Automatic Methodology for Human Action Recognition Incorporating Dynamic Information Fully Automatic Methodology for Human Action Recognition Incorporating Dynamic Information Ana González, Marcos Ortega Hortas, and Manuel G. Penedo University of A Coruña, VARPA group, A Coruña 15071,

More information

Sensor Based Time Series Classification of Body Movement

Sensor Based Time Series Classification of Body Movement Sensor Based Time Series Classification of Body Movement Swapna Philip, Yu Cao*, and Ming Li Department of Computer Science California State University, Fresno Fresno, CA, U.S.A swapna.philip@gmail.com,

More information

Activity recognition and energy expenditure estimation

Activity recognition and energy expenditure estimation Activity recognition and energy expenditure estimation A practical approach with Python WebValley 2015 Bojan Milosevic Scope Goal: Use wearable sensors to estimate energy expenditure during everyday activities

More information

Ubiquitous Computing and Communication Journal (ISSN )

Ubiquitous Computing and Communication Journal (ISSN ) A STRATEGY TO COMPROMISE HANDWRITTEN DOCUMENTS PROCESSING AND RETRIEVING USING ASSOCIATION RULES MINING Prof. Dr. Alaa H. AL-Hamami, Amman Arab University for Graduate Studies, Amman, Jordan, 2011. Alaa_hamami@yahoo.com

More information

AN INTELLIGENT TRAFFIC CONTROLLER BASED ON FUZZY LOGIC

AN INTELLIGENT TRAFFIC CONTROLLER BASED ON FUZZY LOGIC AN INTELLIGENT TRAFFIC CONTROLLER BASED ON FUZZY LOGIC Bilal Ahmed Khan; Nai Shyan Lai Asia Pacific University of Technology and Innovation belalkhn22@gmail.com Abstract Traffic light plays an important

More information

Resistance Spot Welding Process Identification Using an Extended knn Method

Resistance Spot Welding Process Identification Using an Extended knn Method Resistance Spot Welding Process Identification Using an Extended knn Method H. Junno, P. Laurinen, E. Haapalainen, L. Tuovinen and J. Röning Intelligent Systems Group, Department of Electrical and Information

More information

Rail defect diagnosis using smartphones and drones by: Borna Tech.

Rail defect diagnosis using smartphones and drones by: Borna Tech. UIC Digital Awards 2016 Rail defect diagnosis using smartphones and drones by: Borna Tech. Borna Tech 1 Agenda Borna Tech 2 Introduction 2 Introduction Condition monitoring of rail tracks is an important

More information

Finding Dominant Parameters For Fault Diagnosis Of a Single Bearing System Using Back Propagation Neural Network

Finding Dominant Parameters For Fault Diagnosis Of a Single Bearing System Using Back Propagation Neural Network International Journal of Mechanical & Mechatronics Engineering IJMME-IJENS Vol:13 No:01 40 Finding Dominant Parameters For Fault Diagnosis Of a Single Bearing System Using Back Propagation Neural Network

More information

An unconstrained Activity Recognition Method using Smart Phones

An unconstrained Activity Recognition Method using Smart Phones An unconstrained Activity Recognition Method using Smart Phones Naciye Ç elenli, Kamile Nur Seviş, Muhammed F. Esgin, Kemal Altundağ, Umut Uludağ BİLGEM TÜBİTAK (The Scientific and Technological Research

More information

Face Detection CUDA Accelerating

Face Detection CUDA Accelerating Face Detection CUDA Accelerating Jaromír Krpec Department of Computer Science VŠB Technical University Ostrava Ostrava, Czech Republic krpec.jaromir@seznam.cz Martin Němec Department of Computer Science

More information

Keywords APSE: Advanced Preferred Search Engine, Google Android Platform, Search Engine, Click-through data, Location and Content Concepts.

Keywords APSE: Advanced Preferred Search Engine, Google Android Platform, Search Engine, Click-through data, Location and Content Concepts. Volume 5, Issue 3, March 2015 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com Advanced Preferred

More information

A Web Page Segmentation Method by using Headlines to Web Contents as Separators and its Evaluations

A Web Page Segmentation Method by using Headlines to Web Contents as Separators and its Evaluations IJCSNS International Journal of Computer Science and Network Security, VOL.13 No.1, January 2013 1 A Web Page Segmentation Method by using Headlines to Web Contents as Separators and its Evaluations Hiroyuki

More information

Abstract Purpose: Results: Methods: Conclusions:

Abstract Purpose: Results: Methods: Conclusions: Abstract Purpose: Current gaze-tracking systems can monitor point of gaze allowing for free head movement, but do not perform well when used with the large display surfaces and wide range of head positions

More information

Jo-Car2 Autonomous Mode. Path Planning (Cost Matrix Algorithm)

Jo-Car2 Autonomous Mode. Path Planning (Cost Matrix Algorithm) Chapter 8.2 Jo-Car2 Autonomous Mode Path Planning (Cost Matrix Algorithm) Introduction: In order to achieve its mission and reach the GPS goal safely; without crashing into obstacles or leaving the lane,

More information

International Journal of Modern Engineering and Research Technology

International Journal of Modern Engineering and Research Technology Volume 4, Issue 3, July 2017 ISSN: 2348-8565 (Online) International Journal of Modern Engineering and Research Technology Website: http://www.ijmert.org Email: editor.ijmert@gmail.com A Novel Approach

More information

Inertial Navigation Static Calibration

Inertial Navigation Static Calibration INTL JOURNAL OF ELECTRONICS AND TELECOMMUNICATIONS, 2018, VOL. 64, NO. 2, PP. 243 248 Manuscript received December 2, 2017; revised April, 2018. DOI: 10.24425/119518 Inertial Navigation Static Calibration

More information

Time Stamp Detection and Recognition in Video Frames

Time Stamp Detection and Recognition in Video Frames Time Stamp Detection and Recognition in Video Frames Nongluk Covavisaruch and Chetsada Saengpanit Department of Computer Engineering, Chulalongkorn University, Bangkok 10330, Thailand E-mail: nongluk.c@chula.ac.th

More information

Human Activity Recognition in WSN: A Comparative Study

Human Activity Recognition in WSN: A Comparative Study International Journal of Networked and Distributed Computing, Vol. 2, No. 4 (October 2014), 221-230 Human Activity Recognition in WSN: A Comparative Study Muhammad Arshad Awan 1, Zheng Guangbin 1, Cheong-Ghil

More information

Inertial Pen Based Alphabet Recognition using KNN Classifier

Inertial Pen Based Alphabet Recognition using KNN Classifier Inertial Pen Based Alphabet Recognition using KNN Classifier Shaikh J.K. D.Y.Patil College of Engineering, subhedarjj@gmail.com, gulshanashaikh@yahoo.com Abstract In today s electronics world human machine

More information

Character Recognition from Google Street View Images

Character Recognition from Google Street View Images Character Recognition from Google Street View Images Indian Institute of Technology Course Project Report CS365A By Ritesh Kumar (11602) and Srikant Singh (12729) Under the guidance of Professor Amitabha

More information

CS 4518 Mobile and Ubiquitous Computing Smartphone Sensing. Emmanuel Agu

CS 4518 Mobile and Ubiquitous Computing Smartphone Sensing. Emmanuel Agu CS 4518 Mobile and Ubiquitous Computing Smartphone Sensing Emmanuel Agu Smartphone Sensors Typical smartphone sensors today accelerometer, compass, GPS, microphone, camera, proximity Future sensors? Heart

More information

EFFICIENT ADAPTIVE PREPROCESSING WITH DIMENSIONALITY REDUCTION FOR STREAMING DATA

EFFICIENT ADAPTIVE PREPROCESSING WITH DIMENSIONALITY REDUCTION FOR STREAMING DATA INTERNATIONAL JOURNAL OF RESEARCH IN COMPUTER APPLICATIONS AND ROBOTICS ISSN 2320-7345 EFFICIENT ADAPTIVE PREPROCESSING WITH DIMENSIONALITY REDUCTION FOR STREAMING DATA Saranya Vani.M 1, Dr. S. Uma 2,

More information

HAND GESTURE RECOGNITION USING MEMS SENSORS

HAND GESTURE RECOGNITION USING MEMS SENSORS HAND GESTURE RECOGNITION USING MEMS SENSORS S.Kannadhasan 1, R.Suresh 2,M.Shanmuganatham 3 1,2 Lecturer, Department of EEE, Tamilnadu Polytechnic College, Madurai, Tamilnadu, (India) 3 Senior Lecturer,

More information

Semi-Supervised PCA-based Face Recognition Using Self-Training

Semi-Supervised PCA-based Face Recognition Using Self-Training Semi-Supervised PCA-based Face Recognition Using Self-Training Fabio Roli and Gian Luca Marcialis Dept. of Electrical and Electronic Engineering, University of Cagliari Piazza d Armi, 09123 Cagliari, Italy

More information

Realtime Object Recognition Using Decision Tree Learning

Realtime Object Recognition Using Decision Tree Learning Realtime Object Recognition Using Decision Tree Learning Dirk Wilking 1 and Thomas Röfer 2 1 Chair for Computer Science XI, Embedded Software Group, RWTH Aachen wilking@informatik.rwth-aachen.de 2 Center

More information

u Emerging mobile applications that sense context are very important; e.g. for timely news and events, health tracking, social connections, etc.

u Emerging mobile applications that sense context are very important; e.g. for timely news and events, health tracking, social connections, etc. November 11, 2013 Presented by: Merhawit Habte 1. Introduc,on u Emerging mobile applications that sense context are very important; e.g. for timely news and events, health tracking, social connections,

More information

Toward Part-based Document Image Decoding

Toward Part-based Document Image Decoding 2012 10th IAPR International Workshop on Document Analysis Systems Toward Part-based Document Image Decoding Wang Song, Seiichi Uchida Kyushu University, Fukuoka, Japan wangsong@human.ait.kyushu-u.ac.jp,

More information

Wireless Vehicular Blind-Spot Monitoring Method and System Progress Report. Department of Electrical and Computer Engineering University of Manitoba

Wireless Vehicular Blind-Spot Monitoring Method and System Progress Report. Department of Electrical and Computer Engineering University of Manitoba Wireless Vehicular Blind-Spot Monitoring Method and System Progress Report Department of Electrical and Computer Engineering University of Manitoba Prepared by: Chen Liu Xiaodong Xu Faculty Supervisor:

More information

An Efficient Approach for Color Pattern Matching Using Image Mining

An Efficient Approach for Color Pattern Matching Using Image Mining An Efficient Approach for Color Pattern Matching Using Image Mining * Manjot Kaur Navjot Kaur Master of Technology in Computer Science & Engineering, Sri Guru Granth Sahib World University, Fatehgarh Sahib,

More information

An Improved DFSA Anti-collision Algorithm Based on the RFID-based Internet of Vehicles

An Improved DFSA Anti-collision Algorithm Based on the RFID-based Internet of Vehicles 2016 2 nd International Conference on Energy, Materials and Manufacturing Engineering (EMME 2016) ISBN: 978-1-60595-441-7 An Improved DFSA Anti-collision Algorithm Based on the RFID-based Internet of Vehicles

More information

Multimodal Interfaces. Remotroid

Multimodal Interfaces. Remotroid Multimodal Interfaces Remotroid Siavash Bigdeli / Christian Lutz University of Neuchatel and University of Fribourg 1. June 2012 Table of contents 1 Introduction...3 2 Idea of the application...3 3 Device

More information

Improvement of SURF Feature Image Registration Algorithm Based on Cluster Analysis

Improvement of SURF Feature Image Registration Algorithm Based on Cluster Analysis Sensors & Transducers 2014 by IFSA Publishing, S. L. http://www.sensorsportal.com Improvement of SURF Feature Image Registration Algorithm Based on Cluster Analysis 1 Xulin LONG, 1,* Qiang CHEN, 2 Xiaoya

More information

Path Planning with Motion Optimization for Car Body-In-White Industrial Robot Applications

Path Planning with Motion Optimization for Car Body-In-White Industrial Robot Applications Advanced Materials Research Online: 2012-12-13 ISSN: 1662-8985, Vols. 605-607, pp 1595-1599 doi:10.4028/www.scientific.net/amr.605-607.1595 2013 Trans Tech Publications, Switzerland Path Planning with

More information

An Improved Method of Vehicle Driving Cycle Construction: A Case Study of Beijing

An Improved Method of Vehicle Driving Cycle Construction: A Case Study of Beijing International Forum on Energy, Environment and Sustainable Development (IFEESD 206) An Improved Method of Vehicle Driving Cycle Construction: A Case Study of Beijing Zhenpo Wang,a, Yang Li,b, Hao Luo,

More information

Innovative M-Tech projects list

Innovative M-Tech projects list 1 Technologies: Innovative M-Tech projects list 1. ARM-7 TDMI - LPC-2148 2. Image Processing 3. MATLAB Embedded 4. GPRS Mobile internet 5. Touch screen. IEEE 2012-13 papers 6. Global Positioning System

More information

Enhancing Forecasting Performance of Naïve-Bayes Classifiers with Discretization Techniques

Enhancing Forecasting Performance of Naïve-Bayes Classifiers with Discretization Techniques 24 Enhancing Forecasting Performance of Naïve-Bayes Classifiers with Discretization Techniques Enhancing Forecasting Performance of Naïve-Bayes Classifiers with Discretization Techniques Ruxandra PETRE

More information

Cost-sensitive Boosting for Concept Drift

Cost-sensitive Boosting for Concept Drift Cost-sensitive Boosting for Concept Drift Ashok Venkatesan, Narayanan C. Krishnan, Sethuraman Panchanathan Center for Cognitive Ubiquitous Computing, School of Computing, Informatics and Decision Systems

More information

Design av brukergrensesnitt på mobile enheter

Design av brukergrensesnitt på mobile enheter Design av brukergrensesnitt på mobile enheter Tutorial på Yggdrasil Lillehammer, 12 oktober 2009 Erik G. Nilsson SINTEF IKT ICT Agenda 13:15 Introduction, user interfaces on mobile equipment, important

More information

Binary Histogram in Image Classification for Retrieval Purposes

Binary Histogram in Image Classification for Retrieval Purposes Binary Histogram in Image Classification for Retrieval Purposes Iivari Kunttu 1, Leena Lepistö 1, Juhani Rauhamaa 2, and Ari Visa 1 1 Tampere University of Technology Institute of Signal Processing P.

More information

Trip Reconstruction and Transportation Mode Extraction on Low Data Rate GPS Data from Mobile Phone

Trip Reconstruction and Transportation Mode Extraction on Low Data Rate GPS Data from Mobile Phone Trip Reconstruction and Transportation Mode Extraction on Low Data Rate GPS Data from Mobile Phone Apichon Witayangkurn, Teerayut Horanont, Natsumi Ono, Yoshihide Sekimoto and Ryosuke Shibasaki Institute

More information

Terrain Roughness Identification for High-Speed UGVs

Terrain Roughness Identification for High-Speed UGVs Proceedings of the International Conference of Control, Dynamic Systems, and Robotics Ottawa, Ontario, Canada, May 15-16 2014 Paper No. 11 Terrain Roughness Identification for High-Speed UGVs Graeme N.

More information

An Adaptive Threshold LBP Algorithm for Face Recognition

An Adaptive Threshold LBP Algorithm for Face Recognition An Adaptive Threshold LBP Algorithm for Face Recognition Xiaoping Jiang 1, Chuyu Guo 1,*, Hua Zhang 1, and Chenghua Li 1 1 College of Electronics and Information Engineering, Hubei Key Laboratory of Intelligent

More information

Using Association Rules for Better Treatment of Missing Values

Using Association Rules for Better Treatment of Missing Values Using Association Rules for Better Treatment of Missing Values SHARIQ BASHIR, SAAD RAZZAQ, UMER MAQBOOL, SONYA TAHIR, A. RAUF BAIG Department of Computer Science (Machine Intelligence Group) National University

More information

Keywords: wearable system, flexible platform, complex bio-signal, wireless network

Keywords: wearable system, flexible platform, complex bio-signal, wireless network , pp.119-123 http://dx.doi.org/10.14257/astl.2014.51.28 Implementation of Fabric-Type Flexible Platform based Complex Bio-signal Monitoring System for Situational Awareness and Accident Prevention in Special

More information

Fall Detection System for Elderly Based on Android Smartphone

Fall Detection System for Elderly Based on Android Smartphone Fall Detection System for Elderly Based on Android Smartphone Made Liandana, I Wayan Mustika, and Selo Department of Information Technology and Electrical Engineering Universitas Gadjah Mada Jalan Grafika

More information

Object Classification Using Tripod Operators

Object Classification Using Tripod Operators Object Classification Using Tripod Operators David Bonanno, Frank Pipitone, G. Charmaine Gilbreath, Kristen Nock, Carlos A. Font, and Chadwick T. Hawley US Naval Research Laboratory, 4555 Overlook Ave.

More information

ESANN 2013 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges (Belgium), 24-26

ESANN 2013 proceedings, European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. Bruges (Belgium), 24-26 A One-Vs-One Classifier Ensemble with Majority Voting for Activity Recognition B. Romera-Paredes 1,M.S.H.Aung 1, N. Bianchi-Berthouze 1 1 University College London, UCL Interaction Center (UCLIC) WC1E

More information

Estimating Missing Attribute Values Using Dynamically-Ordered Attribute Trees

Estimating Missing Attribute Values Using Dynamically-Ordered Attribute Trees Estimating Missing Attribute Values Using Dynamically-Ordered Attribute Trees Jing Wang Computer Science Department, The University of Iowa jing-wang-1@uiowa.edu W. Nick Street Management Sciences Department,

More information

Minimal Test Cost Feature Selection with Positive Region Constraint

Minimal Test Cost Feature Selection with Positive Region Constraint Minimal Test Cost Feature Selection with Positive Region Constraint Jiabin Liu 1,2,FanMin 2,, Shujiao Liao 2, and William Zhu 2 1 Department of Computer Science, Sichuan University for Nationalities, Kangding

More information

Machine learning based automatic extrinsic calibration of an onboard monocular camera for driving assistance applications on smart mobile devices

Machine learning based automatic extrinsic calibration of an onboard monocular camera for driving assistance applications on smart mobile devices Technical University of Cluj-Napoca Image Processing and Pattern Recognition Research Center www.cv.utcluj.ro Machine learning based automatic extrinsic calibration of an onboard monocular camera for driving

More information

Handheld Person Verification System Using Face Information

Handheld Person Verification System Using Face Information Handheld Person Verification System Using Face Information Yao-Hong Tsai 1, Raymond B.C. Fu 2, Li-Wu Huang, Ching-Chun Huang, and Cheng-Yi Liu Advanced Technology Center, Computer & Communication Research

More information

Customer Clustering using RFM analysis

Customer Clustering using RFM analysis Customer Clustering using RFM analysis VASILIS AGGELIS WINBANK PIRAEUS BANK Athens GREECE AggelisV@winbank.gr DIMITRIS CHRISTODOULAKIS Computer Engineering and Informatics Department University of Patras

More information

Self-Organizing Maps of Web Link Information

Self-Organizing Maps of Web Link Information Self-Organizing Maps of Web Link Information Sami Laakso, Jorma Laaksonen, Markus Koskela, and Erkki Oja Laboratory of Computer and Information Science Helsinki University of Technology P.O. Box 5400,

More information

CLASSIFICATION FOR ROADSIDE OBJECTS BASED ON SIMULATED LASER SCANNING

CLASSIFICATION FOR ROADSIDE OBJECTS BASED ON SIMULATED LASER SCANNING CLASSIFICATION FOR ROADSIDE OBJECTS BASED ON SIMULATED LASER SCANNING Kenta Fukano 1, and Hiroshi Masuda 2 1) Graduate student, Department of Intelligence Mechanical Engineering, The University of Electro-Communications,

More information

Classification of Daily Life Activities by Decision Level Fusion of Inertial Sensor Data

Classification of Daily Life Activities by Decision Level Fusion of Inertial Sensor Data Classification of Daily Life Activities by Decision Level Fusion of Inertial Sensor Data Dominik Schuldhaus, Heike Leutheuser, Bjoern M. Eskofier October 2, 2013 Digital Sports Group Pattern Recognition

More information

Orientation Independent Activity/Gesture Recognition Using Wearable Motion Sensors

Orientation Independent Activity/Gesture Recognition Using Wearable Motion Sensors > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1 Orientation Independent Activity/Gesture Recognition Using Wearable Motion Sensors Jian Wu, Student Member, IEEE

More information