Integrated Circuits and Systems group, IIT Madras

Design and Developmement of a DSP based System for Acquisition and Analyses of EMG Signals

By G. Vijaya Krishna Prasad

Abstract

Electromyography is a method of studying muscle activity by recording voltage changes due to the transmission of signal from the motor end plate of neuron to muscle fibres. The electromyographic (EMG) signal is extremely complicated to interpret since it is affected by properties of muscles, the control scheme of peripheral nervous systems, and the characteristics of instrumentation used to detect and observe it. EMG provides information on the “quality” of contraction, i.e., whether it is continuous, phasic or clonic and also the ability to compare the timing of activity of several muscles. These activities or events are usually dynamic and impossible to view because they occur too quickly or due to the activity of a group of muscles. The measurement of the activity of the muscles (EMG) during working (gait) or other movements, adds a significant information to that obtained by kinetic measurements and is particularly important where a movement disability has a neuromuscular cause such as leprosy, diplegia, hemiplegia, myelo-menigocele and head trauma. The monitoring of activities of muscles during walking is restricted to five muscles in present work. The five muscles of interest are Abductor Hallucis (A.H), Peroneus Longus (P.L), Tibialis Anterior (T.A), Extensor Hallucis Longus (E.H.L) and Gastrocnemius (G.C).

Earlier systems used for acquisition of EMG signals required that a cable, or more specifically, a cable for each channel, physically connected the patient to the data collection device. This entailed the use of heavy and cumbersome multi-cored cables, often with pulley and gantry arrangements to support them while patient walked. Later, the availability of small lightweight frequency modulated (FM) transmitters led to the development of telemetry systems that transmit data from a portable patient unit to a receiving antenna. Since the bandwidth of such FM transmitters are strictly regulated, it results in limiting the ability of the carrier to transmit full frequency, undistorted EMG. Present available commercial EMG systems are used only for the acquisition of EMG and monitoring of muscle activity during gait. The acquired data can not be processed further by the same system in order to extract information from the observed signal. Therefore to acquire, to monitor the muscle activity, and to process the acquired signal when a person is walking with normal or pathological conditions, a DSP based multi channel acquisition and analysis system is designed for use with a personal computer (PC). The idea behind the use of DSP in the system is that the acquired data can be processed further for some other analysis. With the help of designed instrumentation, EMG has been acquired for three normal and three leprosy subjects walking at different velocities. The acquired data has been used for the analysis of EMG activity during walking and for spectral analysis of EMG signals for normal and leprosy subjects.

From the EMG analysis during gait study, it has been observed that the muscles Tibialis Anterior (T.A) and Extensor Hallucis Longus (E.H.L) are active during heel strike phase of walking and the other muscles of interest Gastrocnemius (G.C), Abductor Hallucis (A.H) and Peroneus Longus (P.L) are active in the mid stance and continues through toe off phases of walking for the normal subjects. Similar results are obtained in the case of leprosy subjects provided that the muscle of interest is not paralysed.

While studying high levels of muscle contraction, the EMG signal observed is active only during a particular phase of walking cycle. Short-time energy technique is applied to trace the active segments of the EMG signal. The power spectrum of active segments is estimated using Auto regressive (AR) model of order 5. The integrated power and peak power are determined for each gait cycle and the variations of these parameters with speed is studied. The variations of measurements of the condition velocity ( the mean, the median and the peak power frequency ) are also studied.

It is observed that the G.C has highest peak power followed by A.H, T.A, P.L and E.H.L for normal subjects. Similar trends are observed in the case of leprosy subjects, but with reduced peak power values for non paralysed muscles. The integrated power (total power contributed the motor units in the active state) of each muscle increases with the increase in velocity of walking and also there exists a high degree of correlation (0.95-0.98) between them in normal subjects. Similar trends are observed for the peak power value of the muscles with velocity of walking for the same class of subjects.

For leprosy subjects, the integrated power and peak power have similar variations as normal subjects with velocity of walking for a muscle of grading 5 (completely normal muscle). On the other hand, if the muscle has grading of less than 5 , the peak power and integrated power variations with velocity are very small.

The measures of conduction velocity of muscles (the mean, the median and the peak power frequency) are found to be much lower in the case of normals than those of the leprosy subjects.

In the present work a multi linear regression analysis is done to find the correlation between the normalized foot pressures (found by using barograph) and the EMG activity during different phases of walking cycle, at different velocities of walking for normal and leprosy subjects. The correlation coefficients obtained in the mullet linear regression analysis between the integrated power of the muscles and normalized pressure are having a moderate positive correlation values in normal subjects. Similar results are obtained in the correlation analysis between the peak power and the normalized peak pressure but with still reduced correlation coefficients compared with the earlier case for the same class of subjects. In the case of leprosy subjects, for non paralyzed muscles a higher order correlation coefficients are found than for the normal subjects.