This Small Business Innovation Research (SBIR) Phase II Project proposes the construction of a biosensor device prototype that will produce text from electromyographic (EMG) signals recorded from hand muscles. This biosensor device will enable the user to enter text into a computer or a mobile device without the need of special paper, pen, or other devices to track the pen. Recent advances in various technologies have made it practical to develop the EMG detection and analysis techniques suitable for character recognition. Taking advantage of advances in electrophysiology, pattern recognition, signal processing, and computer engineering, this project proposes to develop a practical system to decipher the EMG signals generated by biosensors mounted in the digital glove. The project will use the test bed system that was developed during Phase I project and helped to prove the concept. The knowledge of hand EMG patterns of various characters that were gained during Phase I will be used in the development of hardware device. The development will be conducted in the areas of Data Collection, Data Representation (preparation), and Data Analysis. The improvements are expected in all three areas, due to the use of more advanced electrodes, data processing filters, and the application of Neural Networks algorithms.
The proposed approach will remove several limitations faced by current technology and should provide a more durable, flexible, accurate, and user friendly product that can be easily adapted to different users for taking notes, or writing SMS messages for cell phones. The technology will significantly impact the condition of Carpal Tunnel Syndrome, a common occupational illness being reported among typists. EMG-based fingerless glove can also be used as alternative communication device by disabled people who are not able to talk, or who have hearing problems. The resulting product has many applications in education, medicine, tele-robotics, and can be used by mobile workers. As a wearable computer device, this product will help to improve users' image and self esteem. This research project will contribute to the better understanding of muscle interactions. Finally, the handwriting application that will be developed, can become a test bed for analyzing and comparing various pattern recognition algorithms, including traditional statistical algorithms and neural networks, for example Self Organizing Maps (SOM), State Vector Machine (SVM), or Time Lagged Recurrent Networks (TLRN). These algorithms already have numerous applications in various fields.
Norconnect Inc PI: Michael Linderman Project number: 0848523 Using Handwriting EMG-EEG signals in Diagnostics of Parkinson’s Disease Project Outcome: We have discovered the biomarkers of Parkinson’s disease. The biomarkers employ a handwriting paradigm and are based on neural data representation in time domain. This system was developed in collaboration with National Instruments, Dartmouth Medical School, St. Lawrence University, Stanford University, and Ogdensburg Free Academy. It is currently at a testing stage in preliminary human trials at Dartmouth-Hitchcock Medical Center. An industrial prototype and the methodology for EMG-EEG and handwriting data acquisition and analysis were developed at Norconnect Inc. Intellectual Merit: In this project we have addressed the problem of Parkinson’s disease early detection. In order to solve this problem we have developed an interface for handwriting paradigm that accurately records the electrical activity from hand muscles (EMG) and Electroencephalographic (EEG) signals from brain. The recorded signals were analyzed and compared in normal and diseased subjects. EMG and EEG signals were processed by computer algorithms developed internally. Handwriting is one of the most important and unique skills developed by human civilization. Neural control of handwriting involves a number of remarkable cognitive and motor mechanisms. We were hoping to detect the distortion in bio-signals at handwriting activity during illness. Our first task was to recognize characters from the large number of hand movements. This was achieved by designing the unique multilayered presentation stage before the signals were passed to a combination of recognition algorithms. The results from this milestone helped us to develop a very convenient digitizing mini-glove which our subjects were wearing together with EEG helmet during the recording. The innovation of our approach was in developing appropriate methods for the analysis of time dependent statistical and correlation properties of neural activity. Broader Impacts: At this time there is no tool on the market that allows for early detection of Parkinson’s disease using comprehensive EEG-EMG analysis. Currently, the examinations are conducted by specialized neurologists and often it takes several years before this disease gets diagnosed. Our light weight solution is simple, inexpensive, has the time resolution of a few milliseconds, and most importantly directly measures the cortex and hand muscle activities during a handwriting task. This test is very sensitive and takes only 40 minutes. Our solution will be used by clinical researchers at universities, hospitals, and medical laboratories. We expect that after the completion of clinical trials this test will be included in yearly medical exams. Publications: "Neuronal variability during handwriting: Lognormal distribution" was accepted for a publication by PLoS ONE. Two patent applications were submitted to USPTO. We are preparing two more papers for academic journals.
