EMG Signal Processing Flow

Figure 4 Physiological Signal Frequency Range The EMG signal is the summation of the discharges of all the motor units within the pick-up range of the electrode. During a sustained weak contraction, a needle electrode might detect two motor units discharging independently at rates of about ten discharges per second. The EMG signal will consist of two distinct trains of motor unit action potentials MUAPs. Most of the MUAPs will be clearly recognizable. Occasionally, the two motor units may discharge at nearly the same time, and the two MUAPs will overlap one another. This condition is called a superposition. The resultant waveform can be either larger or smaller than the individual MUAPs, depending on whether the overlap results in a constructive or destructive interference. Focus is given to surface electromyography sEMG signal due to its potential end-user application in prosthesis or any control mechanism based on muscle signalling. sEMG in health monitoring system is responsible to determine health condition related to muscle activities, in particular musculoskeletal disorder problem. sEMG signal classification also involves muscle fatigue estimation and measurement of muscle recovery rate.

2.3.1 EMG Signal Processing Flow

Figure 5 shows the steps in EMG signal processing. Selection of feature subset with best discrimination ability is still an issue for classifying EMG signals. The success of pattern classification system depends on the choice of features used to represent the raw signals Han-Pang et al., 2003. Even though the raw EMG contains important information of the muscle contraction it needs to be further processed in order to extract its information Konrad, 2005. The filtered raw signal will be further processed with feature extraction which is an essential pre-process step for pattern recognition. Two type of analysis will be performed for the feature extractions which are amplitude and frequency analysis. Figure 5 EMG Signal Processing Flow The large amount of raw EMG data will be processed by performing FFT analysis to acquire the information on the signals because it is impractical to use the raw signal directly for classification as some information might be hidden in the sequence of raw data. It is not valid to directly compare the EMG output of a muscle across subjects because different subjects will have muscles with Inman et al., 1952: • different physiological cross-sections • different lengths - geometry • different ratios of slow- to fast-twitch fibers • different recruitment patterns • different firing frequencies Biosensor Signal Acquisition Preprocessing Feature Extraction Signal Classification Voluntary muscle contraction is the results of communication between the individual muscle fibres of the musculoskeletal system and brain where a thought is transformed into electrical impulses that travels down motor neurons in the peripheral and spine nerves to the neuromuscular junctions that form a motor unit. Individual muscle fibres within each motor unit contract either with all or none response when stimulated which means that the muscle fibre is either contracting to its maximum potential or none at all. Whole muscle contraction strength is dependent on the number of individual fibres which are activated and capable of correlating with the electrical activity measured over the muscle with an EMG sensor.

2.3.2 sEMG Signal Acquisition