Wearable Directional Sound Amplifier. Patrick Zhou, Daehyun Han, Micheal Westfall ECE445 Design Document - Spring 2017 TA : Michael Fatina

Transcription

1 Wearable Directional Sound Amplifier,, ECE445 Design Document - Spring 2017 TA : Michael Fatina

2 Table of Contents 1. Introduction Objectives Background High-level Requirements List Design Block Diagram Microphone Array Power Module Processing Unit Output Physical Design Beamforming and Calculations Flow Chart (Software) Tolerance Analysis Cost and Schedule Cost Analysis Labor Parts Grand Total Schedule Discussion of Ethics and Safety Citations

3 1. Introduction 1.1 Objectives : Microphone arrays are used in many devices today, from hearing aid attachments to more modern commercialized products such as the Amazon Echo. The microphone arrays in these devices are considered small, meaning that they contain fewer than eight microphones. The Amazon Echo contains seven microphones while the Google Home only contains two [1]. A hearing aid with microphone array attachment designed by Bernard Widrow contains six microphones [2]. There has been extensive research done regarding small, uniformly spaced, and fixed microphone arrays but very little regarding large, differently spaced, and mobile arrays. Due to the limited research regarding the benefits of a large (greater than eight microphones) array, Professor Singer asked us to construct one so that he and his team can conduct research using it and determine whether or not large microphone arrays provide benefits over smaller arrays. We intend to develop an easily reproduced and scalable large microphone array that will be worn on the human body. This has the additional benefit of allowing Professor Singer to study the effects of the human body on the sound received by the array. For our own needs, we will be applying a beamforming algorithm onto the real time data we collect from the microphones. This algorithm will amplify sound from specified directions and then the resulting signal will be played through the array wearer s headphones. Microphone arrays such as these can be used to enhance the functionality of commercial products such as hearing aids and virtual reality movies with many additional uses in the future as more research is conducted on them. 1.2 Background : Creating a large, wearable microphone array was specifically requested by Professor Singer. There is currently little to none public research available on wearable microphone arrays. While there are many small microphone arrays in commercial products such as the previously mentioned Amazon Echo and Google Home, these arrays are simply placed at a location within a building. We want to study how the human body affects the sound that the microphones receive; clearly there will be some interference from the body on sounds that are blocked by the body as well as sounds echoing off the body from different directions. 2

4 1.3 High-level Requirement List : Sound within beam amplified, sound outside of beam is attenuated where the width of the beam can be modified through the DSP Enough storage space for 5 minutes of continuous data collection from 8 microphones Be able to detect frequencies up to 8 khz 2. Design 2.1 Block Diagram Figure 1. Block Diagram 3

5 2.1.1 Microphone Array The array will be made of 8 digital microphones that all send their data in parallel to the DSP. They will receive 1.8V from a voltage regulator. Requirements 1. Microphone array picks up sound from 5 +/- 10%m away from a sound source output of 60 +/- 10dB directly in front of the array [5 pts] Verification 1. Have a normal conversation with someone in front of the array and check the output for a signal Battery There will be three AA batteries in series that will provide 4.5 +/-.1V to the voltage regulators. Requirements Verification 1. Stores > 1.2Ah of charge [0 pts] 1. Connect fully-charged battery in a test circuit and discharge the battery at 300mA for 4 hours, making sure voltage remains above 4.5V Processor Unit The unit is made up of a digital signal processor (DSP) with a 3.3V regulator connected to it for its power. The DSP will take the output of the microphone array which is a total of 36 Bytes. Since the array needs to record frequencies of 8 khz, the array needs to sample at the nyquist rate of 16 khz, so the DSP needs to run at least 16 khz. The DSP will apply a filter to the array and send that to the headphones and then rewrite over that memory. Requirements Verification 1. Microphones are sampled within 10% of sampling frequency (16kHz) [15 pts] 1. a. Go to quiet location and record a sound b. Compare output signals from 4

6 each microphone to compare sampling times Output The DSP can not store all the sound information so the DSP will transfer the unfiltered data to flash memory that can store 5 minutes of raw data from each of the 8 microphones. We also must check to ensure that the sound within the beam is amplified and then sound outside is not. Requirements Verification 1. External memory can store 5 minutes of continuous sound from each of the 8 microphones [5 pts] 2. Sound is being amplified [10 pts] 3. Sound within beam is amplified, sound outside is not [10 pts] 4. Dummy data can be upload to FPGA and sent to microcontroller be processed and sent to SD card. [5 pts] a. Play a 5 minute segment of a song near microphone array b. Play stored data on external memory on computer and check to see if the entire song was stored a. Record raw data from a simple, repetitive sound at a fixed location away from the array wearer b. Check the amplitude of frequency response c. Record data after beamforming is applied from the same sound and check amplitude of frequency response d. Second frequency response should have a larger amplitude a. Use two sound sources equidistance from the array wearer, one within the beam and one outside b. Play two distinct but repetitive sounds at equal volumes c. Array wearer should hear the sound within the beam much clearer than the other sound a. The sound will be pre-recorded 5

7 data and be upload via usb link with computer. b. The FPGA will send it to the microcontroller. c. The microcontroller will send the raw data to SD card d. The microcontroller will process it and send process data to analog output (headphones) 2.2 Physical Design Figure 2: Array Microphone Arrangement The physical requirement for the array is to have the largest spatial resolution possible. This means that the array needs the largest aperture available. Since the array is confined to a human s chest, the array should take on the dimensions of it as well; the average human dimensions for the chest is a width of 49 cm and a length of 37 cm, but since smaller people should be able to wear it, 3 cm should be taken off from the width and length to give the final dimensions of a 46 cm by 34 cm configuration. The spacing between each microphone should be equivalent to each other. There will be two spacings, a vertical distance, as well as a horizontal distance. This is because a vertical distance is required to give a 2D spatial resolution for the array. A 2D resolution is required because linear (1D) resolution is limited; it can not distinguish between sound coming from above or below. A 2D resolution allows for the sound to be detected from side to side and up and down. This is important, because if there is an echo in a room and it s coming from the floor or ceiling, the 2D array can attenuate those sounds, while a linear array can not. It is important to note that the more elements in a line the better the resolution as well as 6

8 increased spacing. A rectangle pattern would be ideal, but that would require too many microphones. The next pattern to try is a line going across and a line going down, but this still requires too many microphones.the next idea would be going a diagonal line, but this creates s unintentional problem, the array can t distinguish if sound coming from down or left and up from right, this is because the pattern is symmetrical. Knowing all this an asymmetrical pattern was chosen. Figure 3. Diagram of Array and Memory Worn on Body Above is a diagram of the array as well as the straps used to attach it to the wearer and the separate power and memory storage. The main concerns we had to consider when designing this was keeping the array relatively stable which lead us to use both vertical as well as horizontal straps. The power and memory will be worn on the hip to lighten the load of the array as well as make the device more convenient to wear. 7

9 2.3 Beam-Forming and Calculations An important distinction is that plane waves are the only type wave that is considered for the project and the beam-forming equation. For this to be true a minimum distance is required to establish plane wave assumption. This means that at max sampling rate of 16 khz, the center microphone will detect a frontal plane wave the same time as the outermost microphone. (1) ( y R) 2 + x = R 2 (2) y 2 2 yr + x 2 = 0 x =. 046/2m, y = 343.3m/s/16kHz This gives R to be 1.25m The beam pattern is described below N 1 e i2πfilsin(θ)/c i=0 (3) m agnitude = 20ln(1/N )/ln(10) Where f is the frequency, l is the horizontal or vertical distance between the microphones, and c is the speed of sound. 8

10 Figure 4: Beam Pattern at 8kHz Figure 5: Beam pattern at 1kHz 9

11 2.4 Flow Chart (Software) Figure 6. Top-level Software Flowchart 10

12 2.5 Tolerance Analysis A critical requirement for our project is that all 8 microphones should be sampled at the same time. Using known quantities such as distance from sound source and speed of sound through the median (air at T Temperature), we will calculate the time the sound should take to reach each microphone and then measure the time it takes for the sound to reach each microphone and compare. This can be done by doing direction-of-arrival based beamforming using the array geometry. Any timing discrepancies will throw off the relative delay between samples. The time delay between microphone i and microphone j is shown below, where X i is the distance between microphone i and source, X j is the distance between microphone j and source, c is the speed of sound(340.29m/s), and t ij is the time delay [3]. The more important requirement is that all microphones should be sampled at the same rate. If microphones are all driven by the same clock signal, they will be all perfectly synchronized. However, they could still end up out of sync with each other due to bugs in the code or a processor that can t keep up with incoming data. One way to verify that the channels all match is by making a long recording of a known signal and checking that all the channels line up over the whole recording. 11

13 3. Cost and Schedule 3.1 Cost Analysis Labor Student Hourly Rate Total Hours Invested Total*2.5 $ $14,400 $ $14,400 $ $14,400 Total $43,200 Table 1. Labor Cost Parts For Item Quantity Cost 1.8V Regulator LP5907MFX-1.8/NO P8 1 $ V Regulator AZ1117EH-3.3TRG1 1 $0.43 Microphone Adafruit I2S MEMS Microphone Breakout 15 $ V 1.2Ah Li-ion Battery mAh 1 $9.95 Amplifier MAX $41.7 DSP ADAU1446YSTZ-3 A 1 $

14 Memory AP4GAH322B 1 $8.94 Headphones Molex, LLC $9.95 Total $ Grand Total Section Table 2. Bill of Materials Total Labor $43,200 Parts $ Grand Total $43, Table 3. Total Cost of labor and materials 3.2 Schedule Week Task Responsibility 2/20/2017 2/27/2017 Work on requirements & verifications and physical design on design document Work on tolerance analysis, cost & schedule, and safety concerns Work on editing the block diagram and supporting material Prepare for design review and do research on DSP Prepare for design review and order materials that we need Prepare for design review and do research on MEMS microphone 13

15 3/6/2017 3/13/2017 3/27/2017 4/3/2017 4/10/2017 4/17/2017 Work on power system Work on soldering parts Work on revision of design document Design and submit printed circuit board Design and submit printed circuit board If there are missing parts, talk to the machine shop Work on FPGA Revise printed circuit board Put all together and work on building the circuit Test several arrangements of twelve microphones Run initial test on the microphones Set up the data base Test on the amplified/ attenuated sound Test on the latency of microphones Test on the data storage Prepare for Demos, debugging on the amplified/ attenuated sound Prepare for Demos, debugging on the latency of microphones Prepare for Demos, debugging on the data storage 14

16 4/24/2017 5/1/2017 Final Demonstration & Work on dividing parts on presentation Final Demonstration & Set up a proper demo procedures Final Demonstration & Make sure that all parts work Work on final Paper and check overall grammar Work on final Paper and make sure that everything is returned before checkout Work on final paper and edit presentation slides Table 4. Schedule for each member 15

17 4. Discussion of Ethics and Safety There are three main safety concerns for this projects, making sure that the device does not damage the listener's hearing any farther and not allowing the listener to be injured by the device. Due to the speakers being in the user s ears, it is important to limit the volume of sound coming from them. Human ears can be damaged to long or repeated exposure to sounds reaching 85 db or higher, which can lead to a permanent hearing loss. Therefore, we will limit the outgoing sound to 75 db because sounds of less than 75 db unlikely damage to human ears even for a long time[4]. The microphone array, speakers, and power supply will need to be electrically insulated to ensure that the listener is not accidentally injured by an exposed circuit. Another concern regarding the user s safety is being shocked by exposed wires. We will likely have wires connecting different parts of the product such as from the microphone array placed on the chest to the FPGA and data storage located on the hip. Another safety concern would be with the lithium ion battery. There are inherent risks associated with using a lithium ion battery, particularly when the device is worn by a human. Thermal runaway, where the battery rapidly releases its stored energy, may occur causing the battery to heat up significantly and possibly even cause a fire. These concerns would be mitigated by us designing our circuits regarding the battery more carefully as well as monitoring the temperature of the battery when testing. We will also fill out the appropriate waivers and forms regarding wearable battery packs. The main ethical concern is that people may use this wearable microphone array improperly. Our wearable microphone array should be used to conduct research regarding large wearable microphone arrays but if used incorrectly may be used to potentially spy on other people, thus violating people s privacy and failing to comply with Section 9 of the IEEE Code of Ethics [6]. This ethical guideline states that we should avoid injuring others, their property, reputation, or employment by false or malicious action and we believe that using our array to spy on other parties would violate this guideline. However we are unable to avoid this as amplifying sound specific directions is a critical part of our project. We are going to add a warning label on the product stating that the product should be used for research purposes only. 16

18 5. Citations [1] A. Tilley, "Google home vs. Amazon echo: Everything you need to know," in Forbes, Forbes, [Online]. Available: 56f Accessed: Feb. 15, [2] B. Widrow, "Stanford," in Stanford, [Online]. Available: Accessed: Feb. 15, [3] Tellakula, Ashok Kumar, Acoustic Source Localization Using Time Delay Estimation, in India, [Online]. Available : Accessed : Feb. 24, [4] "Noise-Induced Hearing Loss," [Online]. Available: duced-hearing-loss.pdf. Accessed: Mar. 8, [5] T. R. Long, M. Kahn, and C. Mikolajczak, "Lithium-ion battery hazards," [Online]. Available: Accessed: Feb. 27, [6] (2017) IEEE Code of Ethics. [Online]. Available : [7] 17