DETC ESTIMATING EASE OF SINGLE-FINGERED OPERATIONS OF HAND-HELD APPLIANCES WITH DIGITAL HAND

Transcription

1 Proceedings of the ASME 2011 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference IDETC/CIE 2011 August 29-31, 2011, Washington, DC, USA DETC ESTIMATING EASE OF SINGLE-FINGERED OPERATIONS OF HAND-HELD APPLIANCES WITH DIGITAL HAND Satoshi Kanai Graduate School of Information Science & Technology Hokkaido University Sapporo , Japan Seiya Suzuki Department of Electronics & Information Engineering Hokkaido University Sapporo , Japan ASTRACT Ergonomic-conscious design of hand-held information appliances greatly strengthens their competitiveness. However, current ergonomic assessments are carried out in the form of real user-tests which need many human subjects and expensive physical mockups, and only subjective evaluations are obtained. To solve these problems, in this paper, the Digital Hand where the 3D bone structure, surface skin geometry of the human hand are imitated was developed for quantitatively estimating the ease of single-fingered operations of hand-held appliances. The ease is estimated based on the reachability of a finger tip to several specified operation target s, and on the finger joint angle margin which reflects the mobile range of motions of all operating fingers joints. An effective correlation was found between the estimation measures and the subjective comfort ratings in the fingered operations of a SLR camera. INTRODUCTION With stiffer global market competition of hand-held information appliances such as mobile phones and digital cams, product design aware of ergonomics avoids their commoditization[1] and increases their market competitiveness. There are two aspects of ergonomics in information appliance design; physical and cognitive aspects. The physical ergonomic aspect of information appliance design includes the ease of grasping the housing, manipulating their human interface. However, current ergonomic assessments are carried out in the form of real user-tests which need many human subjects and expensive physical mockups, and only subjective and qualitative evaluations are obtained from the tests. To solve these problems, Digital Hand software, where 3D bone structure, surface skin geometry and finger motion of human hands are modeled with high precision, was developed Digital Hand Generating a virtual grasp & finger 3D model of an appliance s housing & user-interfaces (UIs) Quantitative measures of the ease of fingered operations of UI A plausible grasp for the housing A plausible finger for the UI operation Estimating the ease of fingered operation Figure 1. Overview of estimating the ease of fingered operation using a Digital Hand by our group, and was used for the virtual ergonomic assessment of comfort in grasping products such as cameras, bottles and bicycle parts [2-6]. However, the ease of fingered operations was not evaluated, which is one of the important physical ergonomic measures in operating handheld appliances. The purpose of this paper is to develop a method to quantitatively estimate the ease of fingered operations of handheld appliances using the Digital Hand as shown in Fig.1. The ease of fingered operation is estimated based on the reachability of a finger tip to an operation target and on the finger joint angle margin which reflects the mobile range of the fingers joints. The estimation is integrated with the Digital 1 Copyright 2011 by ASME

2 Digital Hand 3D CAD model of housing Grasp generation method [2-5] Optimum grasp generation Step2 User interaction Selecting UI operating surface Step1 UI operating surfaces Generating operation target s User interaction Specifying operation sequence Collision detection and distance evaluation between a housing & a finger Estimating Reachability Finger joint range of motion Estimating total finger joint angle margin The optimum grasp Operation target s A finger which give the closest distance between a fingertip and a target R S(u) Inverse kinematics by CCD Finger motion generation Pairs of operating finger UI operating surface Finger can / cannot reach the UI The ease of the fingered operation Figure 2. Estimation of the ease of single-fingered operation Step3 Hand software. The correlation was verified between the estimation and subjective comfort evaluations in the fingered operations of a SLR camera. The following sections are organized as follows; first, the previous studies are described. Then, an overview of our Digital Hand software and the optimum grasp generation method are introduced. Then we describe how to generate finger motions for operating user interfaces and how to estimate the ease of fingered operation. Finally, the correlation is examined between the estimation measures and subjective rating in the single-finger operation. RELATED WORKS Recently, researches aimed at modeling a human hand and its applications to ergonomics have been progressing. As for modeling human hand geometry, the Japanese generic hand model was developed by Kouch[7], but the finger motion in the grasp was not simulated. Kurihara[8] proposed a 3D bone structure model from X-ray CT scans, and Miyata[9] also developed a finger joint model from MRIs. ut they did not apply it for the ergonomic assessment. Recently, Shimizu[6] constructed a hand model aimed at ergonomic assessment where geometric accuracies of the 3D finger joint motion in grasping was precisely verified by MRI measurements. Several studies have implemented functions of generating virtual grasp s and for assessing the quality of grasp. Choi[10] proposed an algorithm for generating plausible grasp s. Force closure and grasp quality were used for evaluating fit and comfort of grasp s[2,11] and for selecting plausible grasp s [12]. Database-driven grasp generation and ergonomic assessment for handheld appliances were studied where contact area on the hand model was maximized[5]. Other grasp quality measures based on dynamics of finger motions[13], finger-joint angle distributions[3] and the geometric fitness between finger surface and product shape[4] have also been proposed. Some of them simulated the finger motion, but the ease of fingeredoperation at the grasp had not been studied at all. Miyata[14] studied the ergonomic assessment of fingered operations of cell-phone buttons using their 3D hand model. The operating s used to push buttons were estimated for target s distributed on the button plane through the motion capture data of the hand. The preferable area for buttons was extracted in terms of reachability and margin. However, expensive motion capture data was needed to generate the grasp and the finger motions which becomes a stumbling block to usual design situations. Moreover, they did not simulate the skin surface of the finger, and the collision between the finger shape and the housing shape in operating the buttons was not checked, which may produce plausible evaluation measures even at infeasible finger s. The correlation between subjective comfort and the simulation of the fingered operation was not also fully verified. ESTIMATION PROCEESS OVERVIEW Fig.2 shows the estimation of the ease of single-fingered operations of handheld appliances using the Digital Hand. The process is integrated with the original Digital Hand software[2-5], and mainly consists of the following 3 steps; Step1) Generating the optimum grasp : Our developed methods[2-5] is used for generating the optimum grasp of the Digital Hand for the product housing. A 3D CAD model of the housing is inputted to the generation. In the optimum, the number of contact s[2,4,5] or the fit of the finger surfaces for the product surface edges[3] are automatically maximized. The quality of fit and comfort in the grasp can also be evaluated [2-4]. Step 2) Generating the finger motions: This step consists of 2 sub steps. In the first sub step, operation target s are specified by the user on the 3D CAD model. An operation target is a where a finger tip must reach. In the second sub step, the finger motions which pass through the operation target s are automatically calculated, and the finger joint angles at these target s are estimated using the Digital Hand software. It is assumed that a single-fingered operation is only considered and that the grasp stableness does not degraded during the operation. Step3) Estimating the ease of finger operations: ased on the single-finger motions, the ease of the operations is estimated. Two evaluation measures indicate the ease in this step; reachability and total finger joint angle margin. Reachability measure is a binary index indicating whether a fingertip can physically reach an operation target or not. On the other hand, the total finger joint angle margin is a continuous index, indicating how far the actual joint angles in an operating finger are away from each joint angle limits when a fingertip reaches an operation target in the worst case. 2 Copyright 2011 by ASME

3 Average An initial state The inputted pair of contact s Automatic transformation of the hand (a) Link structure model and Surface skin model (b) Hand size variations Figure 3. The Digital Hand model THE DIGITAL HAND SOFTWARE The Digital Hand software have been developed by our laboratory[2-6]. This software is used as a basis of estimating the ease of fingered operation in this research. The Digital Hand model consists of a link structure model and a surface skin model, as shown in Fig.3(a). The link structure model simulates the rotational motion at each finger joint[2]. The model was constructed based on measurements by MRI[6]. It has 17 links, and each one has a joint on both ends which has 1, 3 or 6 DOF. On the other hand, the surface skin model is composed of a triangular mesh model. The model is able to be deformed using the skin deformation algorithm depending on the finger joint rotation[3]. The digital hand has nine size variations derived from measuring several hundred Japanese adults. Fig.3(b) shows the nine hand size variations based on thickness and length [5]. Two grasp generation methods were prepared in the software in order to obtain the optimum grasp ; a generative method and a database-driven one. In the generative method [2,4], the optimum grasp is generated by a semi-automatic algorithm. ased on a pair of contact s input by the user, the method generates a grasp where the skin surfaces of the hand fits to the product surfaces. The contact area is finally maximized to get the optimum grasp [2]. Moreover, the fit of the finger for the edges of the housing and the margin for the constraints on the finger joint angle limits can be maximized[4]. In the database-driven method in [3,5], the actual grasp s for many samples have been measured using a dataglove to build a grasp database. If a sample shape similar to a given product shape can be found in the database, a nearly appropriate grasp can be obtained. The obtained s may include many infeasible s. To solve the problem, we imposed grasping constraints on the joint range of motion of the human upper limb and the visibility of the fixation area on the synthesis [5]. Under the constraints, a grasp only where the user can actually manipulate the product is left. Finally the was refined so that the contact area becomes maximized. In this research, either the generative or the databasedriven methods were used to obtain the optimum grasp Manual adjustment of the index finger s Automatic finger closing with collision detection The final optimum grasp Figure 4. An example of the optimum grasp generation where the housing is held by five fingers before estimating the ease of fingered operation. Fig.4 shows an example of generating the optimum grasp for a SLR camera using the generative method [4]. GENERATION OF THE FINGER MOTIONS FOR THE OPERATION TARGET POINTS In this process, first, operation target s and the sequence of operation are specified by the user. Second the finger motions which pass through the target s are generated, and the finger joint angles are estimated in the Digital Hand software. Specifying the operation target s An operation target is a on the housing where a finger tip must reach. We only handled the ease of single fingered operations of handheld information appliances in the research. Therefore, the target s must lie on the operating surface of the physical user interfaces such as buttons and dials. The geometries of product housings are represented by triangular mesh models, and the model itself originally does not know which triangles correspond to operating surfaces of the user interfaces. Therefore, first, the user interactively picks up a set of boundary vertices which surrounds an operating surface of the user interfaces on the triangular mesh model, and indicates the user interface type. Currently, a push button type and a dial type can be specified. After picking up the boundary vertices, the operation target s and the normal vectors at the s are automatically extracted from the mesh model according to the user interface type using the following rules. 1) Extracting the target s of a push button As shown in Fig.5(a), first, all triangles bounded by the specified boundary vertices of a push button surface are searched and extracted from the mesh model. Then, a which averages the centroids of all triangles inside the boundary is extracted as an operation target, G, for the push button. Similarly, a normal vector, G, at G is 3 Copyright 2011 by ASME

4 derived as averaged normal vectors of these triangles inside the button boundary. An arrow key has four different operating surfaces, and the boundary vertices of the surfaces must be picked up respectively. As a result, one key can be regarded as four different buttons, each of which has one target. 2) Extracting the target s of a dial On the other hand, three target s are extracted from the operating surface from a dial. The extraction is shown in Fig.5 (b). First, all triangles bounded by the specified boundary vertices of a dial surface are extracted. Secondly, a cylinder whose axis is identical to the dial axis is automatically fitted to the operating surface. Thirdly, all of the boundary vertices are projected to a plane which is perpendicular to the cylinder axis and passes through a centroid of the vertices. Then, a circular arc is fitted to the projected s on the plane. Finally two end s and a mid of the arc are selected as the three target s of D D D G, G and G for a dial. Three normal vectors, D G, D G, D G, each at a target, are also extracted. Specifying the sequence of operation After specifying the operation target s, a sequence of operations also has to be specified by the user. The sequence of operations consists of an ordered combination of an operating surface and an operating fingertip. One of the five finger tips (from thumb to little finger) can be specified as the operating finger tip. [(utton1, Index) (utton2, RingFinger) (Dial3, Thumb) (utton1, Thumb)..] is an example of a sequence of operations. The user can interactively pick the operating surface and the fingertip on the screen of the Digital Hand software. Generating Finger Motions of the Digital Hand ased on the specified operation target s and the sequence of operations, the system automatically generates finger motion of the Digital Hand. This finger motion generation problem can be regarded as the inverse kinematics of an articulated robot manipulator. There are several methods of solving the inverse kinematics problem in robotics[15]. ut these methods lack the flexibility when the end effector cannot exactly reach the target within the joint angle ranges or the D.O.F of the manipulator has redundancy. In this research, a modified Cyclic Coordinate Descent (CCD) method [16] was used for finding the optimum joint angles of the operating finger. The basic idea of the CCD method is performing an iterative heuristic search for each joint angle so that the end effector can reach the target. The method tries to find the best location of the operating finger given its joint angles. This will result in the closest distance of the fingertip and the target. Certain limitations for the joint angles can be imposed in calculating the angles. Fig.6 shows how the inverse kinematics can be iteratively solved using the original CCD method in a 2 dimensional case where a finger has three rotational joints. An outline of solving the process is as follows. J 1 J 2 J 3 D n G 1 n G v G oundary vertices and The operation target the area of the button v G and its normal n G (a) s of a push button oundary vertices and the area of the dial D n G 2 D n G 3 D D D v G 1 v G 2 v G 3 D D D The operation target s v G 1, v G 2, v G 3 and there normals D D D n G 1, n G 2, n G 3 A cylinder fitted to the dial surface (b) s of a dial Figure 5. The extraction process of the target s for the user interface operating surfaces Fingertip J 1 Distal side Proximal side (a) J 1 J 2 (b) (1)Only the finger joint, J 1, at the most distal side is rotated so that the distance between the fingertip and the target is minimized (Fig.6-(b)). (2)The next joint, J 2, is rotated so that the distance is minimized (Fig.6-(c)). (3)The last joint, J 3, at the most proximal side is rotated so that the distance is minimized (Fig.6-(d)). (4)The rotations of (1) to (3) are repeated until the distance between the finger tip and the target becomes lower than a threshold or the distance does not decrease any more (Fig.6-(d)-(h)). During the rotations of (1) to (3), if a joint angle exceeds the joint range of motion, then the angle is fixed to the upper or lower limits. As a result, the optimal finger joint angles (J 1, J 2, J 3 ) can be found which gives the closest distance between the fingertip and the target and satisfies the joints range of motions. However, we must modify the CCD method for the following reason. The surface skin around the fingertip has a rounded shape and consists of many vertices on a triangular J 2 (c) (e) (f) (g) (h) J 3 J 3.. Figure 6. Inverse kinematics process by CCD method (d) 4 Copyright 2011 by ASME

5 For left target of the dial utton #2 For middle target of the dial For middle target of the dial (a) The optimum thumb s for the dial no.12 The optimum (b) The index finger s from the initial to the optimum one for button no.2 Dial #12 utton #11 The optimum (c) The index finger s from the initial to the optimum one for button no.11 Figure 7. Optimum finger s for a dial and two buttons on a SLR camera mesh model. If only one vertex of the mesh model is considered as a fingertip position and the finger is generated only by using this vertex, geometric interference may occur between other vertices around the fingertip and the product housing in the resulting finger. To avoid this interference, we modified the conventional CCD method as follows. A set of vertices located around the fingertip is pre-specified in each finger, and the CCD method is applied between every vertex in this set and the target in order to obtain a set of finger s. Finally, among these finger s, the optimum finger is selected where the angle between the normal vector at the fingertip vertex,, and the one ( G, D G, D G or D G ) at the target vertex becomes closest to. This finger selection does not strictly avoid interference between the fingertip and the housing surface, but practically works well in most cases. Fig.7 shows results of the optimum finger s for a dial and two buttons on a SLR camera housing generated by the modified CCD method. ESTIMATION OF THE EASE OF FIGNER OPERATIONS After finding the optimum finger s for the buttons or dials, the ease of single-fingered operations is estimated under the following two conditions; 1) The user s finger can securely reach the target of the operating surface of the appliance, and 2) The finger should be as comfortable as possible on the operating the surface. Two different measures are introduced for the estimation; reachability measure and total finger joint angle margin. Reachability indicates the degree of first conditions above, while total finger joint angle margin is the second. Reachability Reachability R is a binary measure ( 0,1) indicating whether a finger tip can physically reach a specified target or not. Estimating the reachability consists of two steps. In the first step, we examine whether the operating finger shape at the optimum collides with any portion of the housing shape. As shown in Fig.8, in this test, every vertex,, on the surface skin of the operating finger is first classified into the following three states in the Digital Hand software; If lies outside, then is classified to not collided, where is a set of all vertices in the surface skin mesh model of the Digital Hand, and is the mesh model of the housing shape, if lies inside and its penetration depth,, along the direction of the normal vector,, at is less than a penetration threshold, (Fig.8-(a)), then is classified to contacted, if is not classified to not collided or contacted, then is classified to collided (Fig.8-(b)). ased on this vertex classification, if all vertices including the ones around the fingertip in are then classified to either Finger surface v n v v d v 1 d v τ 1 V H Housing surface M P (a) A vertex v is contacted Fingertip v Finger surface v n v d v 1 d v V H Housing surface τ 1 M P (b) A vertex v is collided Figure 8. Contact test between a vertex on the operating finger and the housing shape tip v G d 2 V H (a) The operating finger can reach the target v G Fingertip vtip d τ 2 d τ 2 d v G 2 V H UI operating surf. Fingertip vertex set (b) The operating finger cannot reach the target v G Figure 9. Reachability test between a vertex on the operating fingertip and the target 5 Copyright 2011 by ASME

6 contacted or not collided, then this finger is examined again in the second step. Otherwise, the is judged to be the one where the operating finger cannot reach the target without colliding the finger body with the housing, and we finally set the reachability R = 0. While in the second step, we examine whether the operating fingertip really reaches the target. As shown in Fig.9, in the optimum finger, if the closest distance,, between a target, G, and a fingertip,, is less than a distance threshold,, then this is judged to be the one where the finger can physically reach the target without any collision of the finger body, and we finally set the reachability R=1. On the contrary, if, the finger is finally judged to be the one where the finger does not collide with the housing but the fingertip cannot physically reach the target, and we finally set the reachability R=0. Total Finger Joint Angle Margin Total finger joint angle margin is a continuous measure ( 0,3) indicating the degree of comfort of the finger when the fingertip touches a button or a dial. This measure evaluates how far the joint angles of an operating finger are away from each joint angle limit in the worst case. We first define a joint angle margin,, of a finger joint, j, when the fingertip comes in contact with a target,, as Eqn(1); 1 where is an angle value of a joint, j, when the fingertip comes into with, and the lower and the upper angle limits of the range of motion of a joint j, and the angles where the active bending of the joint, j, begin to have difficulty in motion. As shown in Fig.10-(a), the joint angle margin takes a trapezoid function of a finger rotation angle. The values of,, and were Joint Angle Margin 1 0 s j ( i ) min j j j Finger Rotation Angle max j j ( i ) (a) The joint angle margin function Figure 10. Joint angle margin MP DIP IP CM (b) Joints in each finger IP MP (1) User interface to be operated The operating finger (a) A finger with large total finger joint angle margin ( S(u) = 3.0 ) The operating finger User interface to be operated (b) A finger with small total finger joint angle margin ( S(u) = 1.4 ) Figure 11. Total finger joint angle margin at different finger s determined by the experiments by human subjects. The details are described in the following subsection. If a user interface,, such as a button or a dial has a set,, of several target s,, then the total finger joint angle margin,, of the is defined as the worst case sum of the joint angle margins,, of the joints belonging to an operating finger by Eqn(2). min (2) where J is a set of finger joints belonging to a operating finger, f, and is defined by Eqn(3).,,,,,,, 3 where, as shown in Fig.10-(b), a is a flexion extension motion of a PIP joint, and and is a flexion extension motion and a forward backward motion of a carpo-metacarpal joint, respectively, an is a flexion extension motion of a DIP joint, and is an adduction and an abduction motion of a metacarpo-phalangeal joint. Fig.11 shows two finger s and their total finger joint angle margin for the different operating surfaces. Measurement of Joint Range of Motion For estimating the total finger joint angle margin, the joint angle limits, and, and the angles, and, at which the bending motion starts receiving reluctance must be determined. These values were obtained from an experiment with humans and the original Digital Hand software as follows. 1) The optimum grasp without operating any interface surface was generated using the Digital Hand software as shown in Fig.12-(a). Also a human subject was asked to copy the grasp of the housing as the optimum one. 2) In the grasp, as shown in Fig.12-(b), the subject was asked to bend a finger one by one until the subject felt the bending limit or until the subject started feeling reluctance in bending, and to stop the finger at these 6 Copyright 2011 by ASME

7 (a) The optimum grasp generated in the Digital Hand (b) Index finger bending test while keeping other fingers in the optimum grasp Joint motion Table 1. Actual joint ranges of motion min j (a) Thumb j j max j IP x CM x CM z (b) Index Joint motion min max j j j j DIP x IP x MP z (c) Feature s (left) and their measurements by 3D CMM (right) (d) A finger reproduced in the Digital Hand by least square fitting Figure 12. Measurement of Joint range of motion s. In the case of the thumb, the subject was asked to move it both in the flexion extension direction and in the forward backward one. While, in the case of other fingers, he/she was asked to move it both in the flexion/extension direction and the adduction/abduction one. 3) As shown in Fig.12-(c), 3D positions of 21 pre-specified feature s placed on the upper skin surface of the subject s hand at the bending s of 2) were measured by a 3D coordinate measuring machine (Micro-Scribe). 4) Using the measuring function proposed by [5], a grasp was then reproduced in the Digital Hand, as shown in Fig.12-(d), by fitting the feature s of the Digital Hand to the measured positions in the lease-square manner. As a result, the values of every finger joint angle at the bending s of 2) can be estimated from the bone link model of the Digital Hand. Four different subjects took part in the experiment. In the Digital Hand software, the hand models with 9 different finger dimensions exist. We selected the one which gave the closest distance between the measured feature s and the ones on the hand model. The average values of,, and among the four subjects were adopted. At this moment, only the joint ranges of motion for a thumb and an index finger were measured. The final results are shown in Table 1. CORRELATION WITH SUJECTIVE RATING Correlation between the proposed measures and the subjective rating of the ease of single-fingered operations was examined in order to validate the measures. Measuring the Subjective Rating The subjective rating of the ease of single-fingered operation of a digital SLR camera (Nikon D80) was measured Figure 13. The user interface surfaces of the camera Figure 14. The physical mockup for subjective evaluation of the camera which had 11 different user interface surfaces including a shutter button, two dials and 1 arrow key, as shown in Fig.13. Eight male subjects whose age ranged from 21 to 24 without any disorder in their fingers participated the measurement. The dimensions of the subjects hands were classified into 9 Digital Hand models as follows; 3 subjects with an average hand model, 1 with a thick hand, 1 with a thin, 2 with long hand, and 1 with a short hand. The measurement was conducted according to the following procedures; 1) A transparent physical mockup of the SLR camera housing shown in Fig.14 made from the 3D mesh model of the camera using a stereolithography machine was used in the single-fingered operation instead of the real camera. 2) The optimum grasp of the housing without operating any the user interface surfaces was generated using the Digital Hand software. 3) The subject was asked to hold the housing mockup with the optimum grasp 4) The subject was then asked to operate the specified user interfaces with the specified finger on the housing one by one. When touching each interface,, the user was asked to answer a subjective rating,, of the fingered operation of each interface. The 4 discrete grade ratings were adopted as ; 0:inoperable, 1:somewhat difficult to operate, 2:operable, 3: very easy to operate. 7 Copyright 2011 by ASME

8 On the other hand, the single-fingered operating is generated in the Digital Hand software, and the reachability R and total finger joint angle margin,, of each interface was estimated. In order to enable correlation of the two estimated measures with the one subjective rating,, to be evaluated, we combined the reachability and total finger joint angle margin into one measure,, using Eqs(4). Average (4) Correlation results Fig.15 shows the results of 0,1,2,3 and 0.0, 3.0 at each user interface no.1-12, respectively. Five different Digital Hand models were used in the estimation according the subjects finger dimensions. From these results, it was found that, at every interface position and in every hand size, the estimated measure,, (red bar) well approximated the subjective comfort rating,, (yellow bar). It was also found that the interfaces of 1, 2, 6 and 12 on the camera housing were placed at the position relevant to the single-fingered operation, while those of 4, 5, 9 and 11 were supposed to be placed at the wrong position and were difficult to operate and should be redesigned if necessary. At this moment, 1.5 can be considered to be the passing condition in ergonomic evaluation for a button or a dial arrangement from the aspect of ease of single-fingered operation. Fig.16 shows the correlation between and when taking all samples. Despite the relatively small number of subjects, a good positive correlation could be found between them, and the correlation coefficient was However, there was still a little large dispersion around the interval, This might be caused by the ambiguity of the subjective feeling of comfort at a high rating. Consequently, an effective correlation was found between the proposed virtual estimation and the subjective comfort ratings in the fingered operations. User Interface No. Thick Thin Long EDH ES Short Figure 16. Correlation between estimated ease of operation ( ) and subjective rating ( ) Figure 15. Estimated ease of operation ( ) and subjective rating ( ) for each user interface surfaces 8 Copyright 2011 by ASME

9 CONCLUSION A method was proposed to quantitatively estimate the ease of single-fingered operations of hand-held appliances using the Digital Hand and the 3D housing model. Ease of the operations was estimated as two quantitative measures; reachability of a finger tip to user interface surfaces, and total finger joint angle margin which reflects the mobile range of motions of all operating fingers joints. The estimation process is automated and integrated with the Digital Hand system. An effective correlation was found between these measures and the subjective ratings of comfort in the single fingered operations of a real SLR camera. A condition of the measures which can be used in judging the relevant placement of user interface surfaces on the housing was also proposed. Of course, many works still remain. The number of subjects in the research was still not enough for conducting general criterion useful in evaluating the relevance of user interface s placements for various single fingered operations and for various appliances. Degradation of the grasp quality during single fingered operations must be considered. In the long term, a muscle-skeletal model of finger motion must be introduced for more reasonable estimation measures which give better agreement with contact force and subjective comfort. ACKNOWLEDGMENTS This research was financially supported by the Japanese MEXT Grant-in-Aid for Scientific Research() under Project No The author would thank to Dr. Yui Endo of the Digital Human Research Center of National Institute of Advanced Industrial Science and Technology for developing the original Digital Hand software and the optimization-based grasp generation algorithm. REFERENCES [1] Christensen, C.M., and Raynor, M.E., 2003, How to avoid commoditization, Harvard usiness School Press, Watertown, MA, U.S.A. [2] Endo, Y., Kanai, S, Miyata, N., Kouchi, M., and Mochimaru, M., 2006, "An Application of a Digital Hand to Ergonomic Assessment of Handheld Information Appliances", SAE Technical Papers, [3] Endo, Y., Kanai, S, Miyata, N., Kouchi, M., and Mochimaru, M., 2007, Virtual ergonomic assessment on handheld products based on virtual grasping by digital hand, SAE Trans. J. of Passenger Cars - Electronic and Electrical Systems, 116(7), pp [4] Endo, Y., Kanai, S, Miyata, N., Kouchi, M., and Mochimaru, M., 2009, "Optimization-ased Grasp Posture Generation Method of Digital Hand for Virtual Ergonomic Assessment", SAE Int. Journal of Passenger Cars -Electronic and Electrical Systems, 1(1), pp [5] Kawaguchi, K., Endo, Y., and Kanai, S., 2009, "Database- Driven Grasp Synthesis and Ergonomic Assessment for Handheld Product Design", Lecture Notes in Computer Science, 5620/2009, pp [6] Shimizu, Y., Kawaguchi, K., and Kanai, S., 2010, "Constructing MRI-based 3D Precise Human Hand Models for Product Ergonomic Assessments", Proceedings of 2010 Asian Conference on Design and Digital Engineering, pp , Aug. 27th, Korea. [7] Kouchi, M., Miyata, N., and Mochimaru, M., 2005, An Analysis of Hand Measurements for obtaining Representative Japanese Hand Models, SAE Trans. J. of Passenger Cars - Electronic and Electrical Systems, 114(7), pp [8] Kurihara, T., Miyata, N., 2004, Modeling deformable human hands from medical images, Proceedings of 2004 Eurographics/ACM SIGGRAPH Symposium on Computer Animation, pp [9] Miyata, N., Kouchi, M., Mochimaru, M., and Kurihara, T., 2005, Finger joint kinematics from MR images, Proceedings of Int. Conf. on Intelligent Robots and Systems, IEEE, pp [10] Choi, J., and Armstrong, T. J., 2006, Examination of a Collision Detection Algorithm for Predicting Grip Posture of Small to Large Cylindrical Handles, Proceedings of 2006 Digital Human Modeling for Design and Engineering Conference, SAE, [11] Pollard, N.S., 2004, Closure and quality equivalence for efficient synthesis of grasps from examples, International Journal of Robotics Research, 23(6), pp [12] Li, Y., Fu, J.L., and Pollard, N.S., 2007, Data Driven Grasp Synthesis using Shape Matching and Task-ased Pruning, IEEE Transactions on Visualization and Computer Graphics, 13(4), pp [13] Yang, J., Pitarch, E.P., Kim, J., and Abdel-Malek, K., 2006, Posture Prediction and Force/Torque Analysis for Human Hands, Proceedings of 2006 Digital Human Modeling for Design and Engineering Conference, SAE, [14] Miyata, N., Kouchi, M., and Mochimaru M., 2006, Posture Estimation for Screening Design Alternatives by DhaibaHand - Cell Phone Operation, Proceedings of 2006 Digital Human Modeling for Design and Engineering Conference, SAE, [15] Paul, R.P., 1981, Robot Manipulators: Mathematics, Programming, and Control, The MIT Press, Cambridge, M.A., U.S.A. [16] Welman, C., 1993, Inverse Kinematics and Geometric Constraints for Articulated Figure Manipulation, Master Thesis, Simon Frasier Univ., urnaby, C, See also URL ftp://fas.sfu.ca/pub/cs/th/1993/chriswelmanmsc.ps.gz. 9 Copyright 2011 by ASME