Method, system and apparatus for real-time classification of muscle signals from self -selected intentional movements

Abstract

A new method, system and apparatus is provided that enables 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. Muscle signals are classified in real-time from self-selected intentional movements. A self-training protocol allows users to select and label their own muscle contractions, and is operable to automatically determine the discernible and repeatable muscle signals generated by the user. A visual display means is used to provide visual feedback to users illustrating the responsiveness of the system to muscle signals generated by the user.

Description

PRIORITY[0001]This application claims the benefit of U.S. patent application Ser. No. 11 / 506,786, filed on Aug. 21, 2006.FIELD OF THE INVENTION[0002]The present invention relates to a method, system and apparatus for classifying muscle signals. More particularly, the present invention relates to a method, system and apparatus for mapping muscle signals to one or more functions of an electronic device.BACKGROUND OF THE INVENTION[0003]References to background documents refer to items on the “List of References” below.[0004]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 [References 1, 2, 3] or from multiple sites. Past studies have employed two [Reference 4, 5, 6], three [References...

AI Technical Summary

Benefits of technology

The system improves user independence and satisfaction by enabling accurate and adaptive control of prosthetic devices based on self-selected movements, reducing the need for extensive user training and healthcare professional involvement.

Problems solved by technology

However, some studies have shown that there is both a practical and theoretical limit to increasing the number of channels [References 10, 5].

It is generally not obvious how previously reported classifiers may be adapted dynamically to accommodate daily variations in EMG signals, due for example, to fatigue.

The main drawback however, of using the transient myoelectric signal as a control input was that it required initiating a contraction from rest.

It severely impeded the coordination of complex tasks involving multiple degrees of freedom.

Therefore, the training of the prosthetic hand and adaptive capability were not automatic.

Unfortunately, the algorithm was not tested on subjects with limb deficiency in real-time.

This presumably would involve significant user training, and therefore Graupe does not provide a method and prosthetic device wherein the training and adaptive capability is automatic.

Apparently, the system could not adaptively change its state over time.

On the other hand, setting a threshold is quite subjective.

Setting this threshold too low makes the system actuation susceptible to noise, and setting it too high, makes the system not responsive to the EMG actuation.

Therefore, the system may not respond to the initiated EMG signals properly and as a result, controlling the powered prostheses, which work based on these algorithms, may become frustrating for the user.

It is also unclear how this method may be beneficial for subjects with limb deficiency in real-time applications.

In U.S. Pat. No. 6,272,479 to Farry et al., the inventors argued that the technical problem common to all the reviewed background art in this area was that no program or system provided a reliable and repeatable output when presented with an input signal or signals having features that were not well characterized or linearly separable or separable on a single feature.

This process is not usually very fast and requires substantial memory and processing resources.

In addition to the shortcomings mentioned above, there are several other general drawbacks to the previous approaches, which are evident to a person skilled in the art.

However, we know that there are significant anatomical and physiological differences between the amputated and intact limb.

In the traumatic amputee, the limb may exhibit scarring, grafted skin areas, muscle atrophy and limited range of motion.

However, these normative modes of activity execution may not be natural for the user and often do not make best use of the individual's abilities.

Consequently, effective device usage requires significant user training.

Unfortunately, setting a threshold is quite subjective.

Setting this threshold too low makes the system actuation susceptible to noise, and setting it too high, makes the system unresponsive to the EMG actuation.

It may be possible to define this threshold by trial and error in a lab, but it is not possible to set this dynamically, as the skin condition (skin perspiration) and other physiological properties of the muscle changes over time, for example, due to fatigue.

Therefore, the system may not respond to the initiated EMG signals properly, and as a result, the use of these devices, for example, powered prostheses, becomes frustrating for the user.

On the other hand, ANN classifiers require a great number of samples to train the network.

There is also a problem of overtraining the network, which may affect the accuracy of the network.

Therefore, these algorithms may not be suitable to be implemented into an embedded system.

It is also not clear how fast these algorithms may adaptively change their inner state in embedded real-time applications, where there are limited resources for memory and processing power.

Unfortunately, there is no report on the successful implementation of these algorithms on a portable (embedded) system.

Images