22
© 2012 Maysam Ghovanloo www.GTBionics.org 1 GT-Bionics Lab School of Electrical and Computer Engineering Maysam Ghovanloo, Ph.D. Dual-Mode Tongue Drive System Using Speech and Tongue Motion to Improve Computer Access for People with Disabilities Xueliang Huo Hangue Park

7.1 - Dual-Mode Tongue Drive System: Using Speech and Tongue Motion to Improve Computer Access for People with Disabilities

Embed Size (px)

DESCRIPTION

Thursday, October 25, 2012 Technical Session #7 Maysam Ghovanloo, PhD – ON Semiconductor Junior Professor, Georgia Tech; Director, GT Bionics Lab; Associate Editor, IEEE Trans. Circuits & Systems II

Citation preview

  • 1. Dual-Mode Tongue Drive System Using Speech and Tongue Motion to Improve Computer Access for People with Disabilities Xueliang Huo Hangue Park Maysam Ghovanloo, Ph.D. GT-Bionics Lab School of Electrical and Computer Engineering 2012 Maysam Ghovanloo 1www.GTBionics.org
  • 2. Some Events and Their Consequences 2012 Maysam Ghovanloo 2www.GTBionics.org
  • 3. Spinal Cord Injuries (SCI) 2012 Maysam Ghovanloo 3www.GTBionics.org
  • 4. 54M Americans (~20%) 11,000 cases of severe SCI add every living with disabilities year to a total population of 55% 250,000 of the SCI victims are They need lifelong 16~30 special care Financial, emotional, and productivity years old cost to the families and the society 2012 Maysam Ghovanloo 4www.GTBionics.org
  • 5. How to Improve their Quality of Life? 2012 Maysam Ghovanloo 5www.GTBionics.org
  • 6. Why Using Tongue? Along with mouth occupies the amount of sensory and motor cortex that rivals fingers and hand: sophisticated motor control capability evident in speech and ingestion Fast movement with many degrees of freedom (DoF). Very flexible Connected to brain by a cranial nerve: escapes even high level spinal cord injuries Noninvasive access to tongue is possible. Motor Homunculus Not afflicted by repetitive motion disorders Does not fatigue easily. Very low rate of perceived exertion Cosmetic advantage and privacy. It is all inside the mouth Not influenced by the position of the rest of the body Hypoglossal Unlike BCIs does not need concentration. Nerve 2012 Maysam Ghovanloo Ghovanloo, IEEE EMBC 2007 6www.GTBionics.org
  • 7. Dual-Mode Tongue Drive System (dTDS) 3-axial magnetic sensors + microphones on headset Wireless data Processing in smartphone Wireless commands Target devices 2012 Maysam Ghovanloo US Patents 8044766, 8242880, Other patents pending 7www.GTBionics.org
  • 8. Magnetic Tracer Attachment Temporary attachment: By tissue adhesives to test-drive the TDS Semi-permanent attachment: By magnetic tongue piercing to use the TDS on a regular basis 2012 Maysam Ghovanloo 8www.GTBionics.org
  • 9. Command Definition and Training Current version: Six tongue commands plus the tongue resting position as neutral 2012 Maysam Ghovanloo 9www.GTBionics.org
  • 10. TDS Clinical Trials Computer access and wheelchair navigation using TDS 2012 Maysam Ghovanloo 10www.GTBionics.org
  • 11. dTDS Wireless Headset 2012 Maysam Ghovanloo 11www.GTBionics.org
  • 12. dTDS Specifications Specification Value Magnetic Sensors Type Honeywell HMC1043 AMR sensor Dimensions 3.0 3.0 1.5 mm3 Sensitivity / range 1 mV/V/Gauss / 600 T Microphone Type SiSonic SPM0408HE5H Dimensions 4.7 3.8 1.1 mm3 Sensitivity / SNR -22 dB / 59 dB Control Unit Microcontroller TI CC2510 SoC Wireless frequency / data rate 2.42 GHz / 500 kbps Sampling rate 50 sample/s/sensor Number of sensors /duty 4 / 8% cycle Audio codec / interface TLV320AIC3204 / I2S Audio sampling rate / 8 ksps / 16 bits / -Law resolution / compression 3 V / 35 mA (audio on) Operating voltage / current 6 mA (audio off) PCB Dimensions 36 16 mm2 Wireless data transfer to Headset Material Object VeroGray resin a PC running Dragon Total weight 90 g (including battery) Naturally Speaking 2012 Maysam Ghovanloo 12www.GTBionics.org
  • 13. dTDS Performance Evaluation 2012 Maysam Ghovanloo 13www.GTBionics.org
  • 14. Experimental Design 14 able-bodied subjects (age: 21 30 years old, 9 males and 5 females) from the Georgia Tech graduate and undergraduate student population 7 subjects had prior experience with TDS, and 7 were naive 7 subjects were native English speakers and 7 were non-native A within-subject model with each subject repeating the same tasks using three devices: TDS, Dragon, and dTDS Two sessions, ~3h each, maximum of one week apart: 1) Instructional session: Subjects learned to use TDS, Dragon, and dTDS 2) Experimental session: Quantitative and qualitative measurements Two tasks: 1) Text transcription: subjects transcribed two short paragraphs from a hard copy onto a word document to test acoustic input 2) Maze navigation: subjects navigated the mouse cursor through an on- screen maze, clicked on designated areas, and typed words/numbers Questionnaire: At the end subjects rated their experience with each device 2012 Maysam Ghovanloo 14www.GTBionics.org
  • 15. Experimental Methods R L R R R L R L L R L Subjects were asked to go through the maze as quickly and accurately as possible, issue right/left click on yellow/green boxes, followed by typing A minimum of 12 cursor movements, 11 clicks (excluding those for typing), and typing 36 characters (on average) in each round of the trial. Subjects were also asked to perform both tasks with a combination of standard mouse and keyboard to generate a reference point. Performance Measures: Recognition accuracy, total completion time, typing time, navigation time, typing error, and navigation error (deviation) 2012 Maysam Ghovanloo 15www.GTBionics.org
  • 16. Experimental Results Total completion time: Significant effect of device on all three measures Recognition accuracy: Microphone type Native Speakers Non-native Speakers Significant (7)* (7)* effect of Commercial 94.0% 1.9% 78.6% 5.9% accent dTDS 91.5% 2.3% 75.7% 5.1% 2012 Maysam Ghovanloo 16www.GTBionics.org
  • 17. Qualitative Results Recognition accuracy: Novice subjects only (7) A higher number represents more positive perception 2012 Maysam Ghovanloo 17www.GTBionics.org
  • 18. Intraoral Tongue Drive System (iTDS) 2012 Maysam Ghovanloo Park and Ghovanloo, ISSCC 12 18www.GTBionics.org
  • 19. iTDS Implementation S3 S1 MSP 25mm 430 Front 20mm iTDS chip S2 S4 35mm Magnetoresistive Sensors 49mm[ Top view ] [ Bottom view ] 42mm Charging coil Tx Li-Ion @13.56MHz Antenna rechargeable 432MHz Battery 2012 Maysam Ghovanloo Park and Ghovanloo, ISSCC 12 19www.GTBionics.org
  • 20. Conclusions Tongue Drive System (TDS) is a wireless, wearable, and minimally invasive brain-tongue-computer interface (BTCI) that enables individuals with severe disabilities to access and control with their voluntary tongue motion. The latest dual-mode TDS (dTDS) prototype appears as a wireless headphone with both tongue motion and speech recognition (SR) capabilities for navigation and typing, respectively. The subjects performance with the dTDS was significantly better than unimodal TDS and SR in a task that involved both navigation and typing (e.g. web surfing). Subjects preferred dTDS over either TDS or SR in terms of speed, ease of use and overall satisfaction. We are now in the process of developing a multimodal Tongue Drive System (mTDS) as well as an intraoral Tongue Drive System (iTDS) that is completely inconspicuous. 2012 Maysam Ghovanloo 20www.GTBionics.org
  • 21. GT- Bionics Lab Members 2012 Summer picnic 2012 Maysam Ghovanloo 21www.GTBionics.org
  • 22. Acknowledgements Funding provided by: National Science Foundation Christopher and Dana Reeve Foundation National Institutes of Health Army Research Office (ARO) ON Semiconductor Collaborators: Dr. Michael Jones, Shepherd Center, Atlanta, GA Dr. Ann Laumann, Northwestern University, Chicago, IL Dr. Joseph Manns, Emory University, Atlanta, GA Dr. Elliot Roth, Rehab. Institute of Chicago, Chicago, IL Dr. Elizabeth Bailey, University of Arizona, Tucson, AZ Dr. Karim Oweiss, Michigan State University, Lansing, MI Dr. Kimberly Wilson, Emory Hospital, Atlanta, GA Dr. Stephen Sprigle, Georgia Tech, Atlanta, GA 2012 Maysam Ghovanloo 22www.GTBionics.org