Facial EMG signal analysis method and its implementation as a stand-alone system

Abstract

This paper describes the second generation of a facial muscle-activity (electromyography –EMG) based communication system.  The previous generation system has been successfully used to help people with physical disabilities control devices such as nurse-call equipment, computers and other peripheral devices.  Our new device is much easier to use than existing commercial systems, does not require the use of a computer, is portable and has a battery life of at least 36 hours. It consists of a headband, a connecting cable and a signal processor that samples and analyzes the s-emg signals. The headband holds three surface-electromyography (s-emg) electrodes that constitute a single channel of differential s-emg data. The signal processor employs novel wavelet-based routines using only logical operations and summations on a low-power 8-bit microcontroller. The use of wavelets significantly reduces costs (through the employment of low cost microprocessors) and, compared to Fast Fourier Transform (FFT) processing greatly increases the system’s ability to detect desired signals and reject artifacts and noise. We have undertaken clinical trials that have demonstrated the system’s capability of rejecting background noise and artifacts from eye-blinks and other involuntary facial movements, while at the same time reliably detecting eyebrow-raises and jaw clenches. We believe that our system could form the basis of very robust, reliable and low-cost communication and environmental control system for subjects with severe physical challenges, arising from, for example, ALS, Parkinson disease, muscular dystrophy and multiple sclerosis.