Lips-Control Assistive System for Communication Presenter: Wei-Min Chang Advisor: Dr. Shih-Chung Chen Date: 2006/10/25 Bio-Medical Electronic Center. Institute of Electrical Engineering Southern Taiwan University of Technology 1 Outline 1. Motivation 2. Purpose 3. Paper Review 4. Materials and Methods 5. Experimental Results 6. Conclusions 7. Future Works 8. References 2 Motivation • Many computers’ input devices are designed for normal persons, so these devices are unsuitable for the disables. • Many researchers develop many auxiliary devices for the disables ,but these auxiliary devices still have many defects when the disables use them in real life. 3 Purpose • We hope to realize a set of lips-control assistive system for communication by labview without wearing any auxiliary devices for the disables with cerebral palsy or basket case. • “Morse Code-Based Mouth Controlled Input Device with Fuzzy Recognition for the Severe Spinal Cord Injuries” had developed a stonking communication system by mouth-control. 4 Paper Review • Pham The Bao, Jin Young Kim, Seung You Na, “FAST MULTI-FACE DETECTION IN COLOR IMAGES USING FUZZY LOGIC”, Proceedings of 2005 International Symposium on Intelligent Signal Processing and Communication Systems. 5 Paper Review • Anagnostopoulos C., Anagnostopoulos I. ,Vergados D. , Papaleonidopoulos I. , Kayafas E., Loumos V. and G. Stasinopoulos,” A Probabilistic Neural Network for face detection on segmented skin areas based on fuzzy rules”, IEEE MELECON 2002, May 7-9,2002, Cairo, EGYPT. 3 Paper Review • Ching-Hsing Luo, Chung-Min Wu, Shu-Wen Lin, Tsan-Hsun Huang, Cheng-Hong Yang, Ming-Che Hsieh, Shih-Chung Chen and Chih-Kuo Liang, “Mouth-Controlled Text Input Device with Sliding Fuzzy Algorithm for Individuals with Disabilities”, IEEE instruement and measurement 2005 (submitted). • Chung-Min Wu and Ching-Hsing Luo, “Morse code recognition system with fuzzy algorithm for disabled persons” Journal of Medical Engineering & Technology 2002, Vol 26 • Wu, C. M., Luo, C. H., Lin, S. W., Chen, S. C., Hsieh, M. C., Chao, C. T., & Tai, C. C. (2002). “Morse code recognition system with adaptive fuzzy algorithm for the disabled”. Journal of Medical and Biological Engineering, 22(4), 205-210. …… Fuzzy Algorithm Controller Goal Input of Hardware (morse code) 4 Materials and Methods • Equipment of Experiments a. Computer : intel pentium4 2.4G b. SDRAM 512mb c. High resolution digital color ccd : CH-926 HK 1/3” d. Color image acquisition card : NI PCI-1411 e. Format of image transmission : NTSC • Software of Experiments LabVIEW 7.1 Vision assistant 7.0 LabVIEW PID toolkit 4 Materials and Methods • Schematic Diagram of System Structure (New (Old Version) Version) ( Image processing algorithm ) Image processing algorithm user Parallel port ccd McTin PC Ps2 morse code Morse code System algorithm utilizing fuzzy theory ( Fuzzy algorithm, Calling API) Database functions 4 Materials and Methods • Technology of Kernel Processing • 1. Image Processing Acquiring images from CCD (RGB) Face tracking and detection algorithm Allocation and extraction of lips images Processing and recognition of lips images 2 Materials and Methods • Technology of Kernel Processing • 2. Fuzzy Theory • Conventional or crisp sets are binary. An element either belongs to the set or doesn't. • A question of a degree of association. Everything is a matter of degree ! 2 Materials and Methods • Technology of Kernel Processing • 2. Morse Code Communication Protocol i.e. Character “A” dot space dash space Next Character n(k) dot-dash space Protocol: Character space (1) dot space:dash space=1:3 (Tone space) (2) dot-dash space:character space=1:3 (Silent space) 2 Materials and Methods • Technology of Kernel Processing • 2. Morse Code Communication Protocol t t PB t t PS Long sound NM Threshold Short sound 2 Materials and Methods • Technology of Kernel Processing • 2. Fuzzy algorithm Applied in Mctin System fT Xk + ek - Fuzzy Algorithm yk + yk-1 *2 Ik e’k + Z-1 Tk Controller 2 Materials and Methods • Technology of Kernel Processing • 3. Windows API • The Microsoft Windows application programming interface (API) provides building blocks used by applications written for Windows . • You can provide your application with a graphical user interface; display graphics and formatted text; and manage system objects such as memory, files, and processes. 2 Materials and Methods • Technology of Kernel Processing • 3. Windows API Open notepad file Send? Yes Paste End No Start keybd_event Send FindWindowA SetWindowPos Notepad file open? Yes No Paste End Auto open notepad file 2 Example 6 Materials and Methods • Verification of fuzzy recognition algorithm by software Test1: Data of expert Taking the data analysis for experts in reference[7] . Experimental method is ditto . 使用參考文獻[7]中所使用的專家數據。實驗方式同上 Test2: Data of a cerebral palsy Taking the data analysis for a teenager with cerebral palsy in reference[7] . Experimental method is that random reading that 20 data, 200 numbers every time, ten times in operation . 使用參考文獻[4]中所使用的一名十幾歲的腦性麻痺患者數據。實驗方式為隨機讀取這20筆資料,每次共200筆數據,共進行 十次 • Verification of fuzzy Test3: Data of normal recognition algorithm by human 20 20~30 year-old normal person who input long-short sound data array which divided into “long” 、”short” 、”long” 、”short”….Number of data is 100. (Long and short sounds are 50 and 50 respectively ). 由二十位20~30歲的正常人,以按鍵的方式,每人輸入固定的長短音資料組,分別為一長、一短、一長、一短…共100筆資料 (長、短音各50筆) Test4: Fuzzy1 & fuzzy2 One disable with Spinal injury who input long-short sound data array which divided into “long” 、”short” 、”long” 、”short”….by open and close mouth. Number of data is 20. (Long 4 Experimental Results • Verification of fuzzy algorithm by software Test1.Testing of the data analysis for experts nth Number of right/data correct rate (%) Dash Dot Lm (ms) Sm (ms) 1 197/200 98.50 1 249 63 2 198/200 99.00 2 274 72 3 197/200 98.50 3 263 99 4 196/200 98.00 4 304 57 5 198/200 99.00 5 259 92 6 197/200 98.50 6 318 61 7 191/200 95.50 7 365 102 8 197/200 98.50 8 197 60 9 200/200 100.00 9 256 68 10 200/200 100.00 10 394 114 Average ± sd No. 98.55±1.2 6 The data analysis for experts Result of experiment of the data analysis for experts 6 Experimental Results • Verification of fuzzy algorithm by software Test2. Testing of the data analysis for a teenager with cerebral palsy Dash Dot Lm(ms) Sm(ms) 1 619 163 2 645 117 3 677 110 4 812 79 5 733 75 6 634 73 7 755 125 8 969 139 9 749 134 10 1238 146 No. nth Number of right/data correct rate (%) 1 198/200 99.00 2 197/200 98.50 3 198/200 99.00 4 199/200 99.50 5 198/200 99.00 6 197/200 98.50 7 193/200 96.50 8 198/200 99.00 9 200/200 100.00 10 199/200 99.50 Average ± sd 98.85±0.9 4 The data analysis for a teenager Result of experiment The data analysis for a teenager with cerebral palsy with cerebral palsy 6 Experimental Results • Verification of fuzzy algorithm by human Test3.Testing of normal ( FUZZY1) nth Number of right/data The longto-short ratio ± sd correct rate (%) 1 97/100 3.48±0.88 97.00 2 98/100 6.59±2.18 98.00 3 94/100 2.61±0.76 94.00 4 99/100 3.85±0.87 99.00 5 97/100 6.71±1.73 97.00 6 97/100 3.46±0.74 97.00 7 96/100 7.27±2.59 96.00 8 93/100 8.34±4.48 93.00 9 98/100 3.75±0.68 98.00 10 98/100 4.89±0.98 98.00 11 98/100 3.34±0.46 98.00 12 97/100 5.58±2.21 97.00 13 98/100 2.51±0.49 98.00 14 95/100 4.24±0.96 95.00 15 84/100 2.50±1.26 84.00 16 94/100 3.47±0.96 94.00 17 99/100 4.40±0.92 99.00 18 55/100 1.58±0.31 55.00 19 97/100 3.67±0.97 97.00 20 95/100 3.52±0.74 95.00 Average ± sd 93.95 ±9.74 Result of experiment of normal 6 Experimental Results • Verification of fuzzy algorithm by human Test4.Testing of disables (FUZZY1) Fuzzy1 nth Number of right/data The longto-short ratio ± sd correct rate (%) 1 17/20 3.65 ±1.74 85.00 2 18/20 3.83 ±2.26 90.00 Avera ge ± sd 87.5 ±3.54 x-axis :Number y-axis :Time (ms) 6 Experimental Results • Verification of fuzzy algorithm by human Test4.Testing of disables (FUZZY2) Fuzzy2 nth Number of right/data The longto-short ratio ± sd correct rate (%) 1 19/20 3.47 ±1.56 95.00 2 19/20 3.54 ±1.54 95.00 Avera ge ± sd 95 .0 ±0.00 x-axis :Number y-axis :Time (ms) 6 Conclusions • Case 18 • The effect of system performance by adding fuzzy algorithm • The effect of light on system. • Fuzzy1 & fuzzy2 6 Future Works • Testing for the disables. • To modify image processing algorithm • Function test of the remote control system of electrical home appliances. • Mouse control • Testing for the effect of system performance by adding fuzzy algorithm 17 References • “ 認識fuzzy-第二版,” 王文俊, 全華科技, 2001 • LabVIEWTM PID Control Toolset User Manual • LabVIEWTM Fuzzy Logic for G Toolkit Reference Manual • “LabVIEW & Microsoft 的整合實例(I),” 陸光中/蕭子健, 高立圖書, 2004 • LabVIEWTM Using External Code in LabVIEW User Manual • Ching-Hsing Luo, Chung-Min Wu, Shu-Wen Lin, Tsan-Hsun Huang, Cheng-Hong Yang, Ming-Che Hsieh, Shih-Chung Chen and Chih-Kuo Liang, “Mouth-Controlled Text Input Device with Sliding Fuzzy Algorithm for Individuals with Disabilities”, IEEE instruement and measurement 2005 (submitted). • Chung-Min Wu and Ching-Hsing Luo, “Morse code recognition system with fuzzy algorithm for disabled persons” Journal of Medical Engineering & Technology 2002, Vol 26 • “應用於重度脊髓損傷患者之摩斯碼模糊辨識嘴控輸入系統,” 國立成功大學電機系, 吳崇民, 博 士論文, 2004 18 Thanks for Your Attention!! 19 Results of Experiments Performance testing of image processing and fuzzy algorithm Test3. performance testing Data of expert Data of cerebral palsy Image processing Number Average (ms) Sd (ms) 105 24.03 7.73 Number Average (ms) Sd (ms) 105 23.13 8.72 Number (Frame) Average (ms) Sd (ms) 105 271.87 23.63 6 Results of Experiments Timing diagram Image processing Status of lip Processing of fuzzy algorithm 6 Results of Experiments Verification of image recognition algorithm by human Test3.正常人數據測試 專 敘述統計 個數 104 104 VAR001 有效的 N (完全排除) 平均數 .02403 標準差 .007731 敘述統計 VAR00 1 有效的 N (完全排除) 個數 10 6 10 6 平均數 .0 23 13 標準差 .0 08 71 9 p 敘述統計 VAR001 有效的 N (完全排除) 個數 104 104 平均數 271.86538 標準差 23.632731 敘述統計 VAR001 有效的 N (完全排除) 個數 106 106 平均數 .02109 標準差 .008880 影 影自 輸 表6. 正常人數據測試實驗結果 6