INVESTIGATION OF FACTORS AFFECTING COLLISION CVD ESTIMATION AND THE IMPACT OF DECOMPOSITION ERRORS ON THE EMG SIGNAL COHERENCE
Majeti, Srivatsa Subba Rao
MetadataShow full item record
Experimental measurements are never perfect, even with sophisticated modern instruments. One of the fundamental problems in signal measurement is distinguishing the noise from the signal. Sometimes the two can be partly distinguished on the basis of frequency components: for example, the signal may contain mostly low-frequency components and the noise may be located at higher frequencies. This is the basis of filtering. This thesis discusses some changes in the experimental protocol such as determining a suitable stimulation site to elicit full compound nerve action potentials (CNAP). The effect of sampling frequency and smoothing techniques to improve the resolution of the conduction velocity distribution (CVD) estimates will also be discussed. A change in stimulation site to elicit the full CNAPs is proposed after realizing that it is relatively difficult to stimulate at the same location to recruit the nerve fibers repeatedly at the elbow. Thus, the stimulation site was changed from elbow to wrist to elicit the full CNAPs. From the simulations it is evident that there was some signal information beyond 2.5 kHz frequency resulting in an increase in the sampling rate from 5 kHz to 10 kHz. The results obtained after employing smoothing techniques improved the CVD resolution. The simulation results were corroborated with the experimental results obtained. Another aspect of this thesis is to check the error tolerance of the EMG decomposition algorithm. Once the muscle electrical activity is recorded, MU trains undergo an automatic decomposition process. Decomposition errors are present in most contractions, thus a human operator has to make changes/correct the values of the motor unit firing times. From the data acquired, false negatives, false positives and false negative-positive errors have been introduced. Different levels of errors to measure the coherence between two motor-unit firing trains from a muscle contraction were also introduced. The firing rate curves are computed for each MU to analyze the interactions between two motor units (MU). The false negatives type of errors was found to be least detrimental. Whereas the false positives and false negative-positive errors affected coherence the most, their error tolerance was only a single error per 5 seconds.