Method, System and Apparatus for Real-Time Classification of Muscle Signals from Self-Selected Intentional Movements
Externally powered prosthetic hands are typically controlled using electromyographic (EMG) signals. These signals originate from the polarization and depolarization of the muscle membrane during voluntary contractions and can be measured at the skin surface using either dry or wet-type electrodes. The EMG control signal can be derived from a single site or from multiple sites. Past studies have employed two up to eight recording sites with varying levels of success. However, some studies have shown that there is both a practical and theoretical limit to increasing the number of channels.The Opportunity
A much sought after goal in EMG-driven prostheses is to provide the user with multiple limb functions, such as hand opening and closing, and wrist rotation. To control multiple functions, it is necessary to map EMG signals corresponding to different muscle contractions to a variety of prosthetic functions. This mapping is commonly achieved by way of a signal classification scheme. In the past decade, many different EMG classification schemes have been proposed for prosthesis control. Rather than requesting users to perform activities according to norms for executing a target activity (which may not apply, or may apply to a varying degree, to particular users), or to generate a muscle signal that meets a particular threshold (which may not apply, or may apply to a varying degree, to particular users), the present invention operates on the basis of analysis of a range of discernable and reproducible muscle signals that a particular user is capable of generating.
The Solution
Signals, recorded from 8 sites on an amputee and an able-bodied subject, could classify six movements out of seven with 100% accuracy. It was found that KNC network comprising 25 nodes was sufficient to classify hand movements. This network training was completed after 500 repetitions. This would enable muscle signals that correspond to muscle contractions to be mapped to one or more functions of an electronic device such as a prosthetic device or gaming apparatus. Further, muscle signals are classified in real-time from self-selected intentional movements.
The Technology
The muscle signals are then processed to extract unique features from the signals, which can be accomplished using a suitable feature extractor means. In a preferred embodiment of the present invention, the features are extracted by calculating the natural logarithm of root-mean-square values. These features then define a feature space, which can be subsequently clustered using a suitable clustering algorithm.
Type of Offer: Licensing
« More Software Patents