The effect of electrode arrangement on spectral distance measures for discrimination of EMG signals.

The proper electrode placement in applying cepstral coefficients for electromyogram (EMG) signature discrimination was investigated. We measured EMG signals of different motions with two electrode arrangements simultaneously. Electrode pairs were located separately on dominant muscles (S-type arrangement) and closely in the region between muscles (C-type arrangement). The application of the cepstral method to signals derived from a C-type arrangement did not achieve the same discrimination as with a S-type arrangement. We used a simplified model to elucidate the poor performance in C-type signals. The bandwidth of signals obtained from S-type placement is wider than that from C-type. Narrower bandwidth decreases the importance of the more discriminative parts for both autoregressive (AR) and cepstral methods. The cepstral method is more sensitive to such variation, so the degradation in performance is more serious for the cepstral method. Second, the amplitude of C-type signal is lower than the S-type; therefore, the C-type signal is more sensitive to the disturbance of noise, especially in the high-frequency band. As high-frequency noise increases, the spectral difference between different EMG signals is gradually dominated by the low-frequency part, which is more informative. Thus, the performances of both methods are improved with increasing high-frequency noise. The improving rate of the AR method is faster than the cepstral method; therefore, its discriminative efficiency may exceed the cepstral method with C-type arrangement.

A comparative analysis of various EMG pattern recognition methods.

Identification of motions of the neck and shoulders using the electromyographic (EMG) signal was investigated in this study. Three discrimination methods, the Euclidean distance measure (EDM), the weighted distance measure (WDM) and the modified maximum likelihood method (MMLM), were used to compare the conventional autoregressive (AR) and cepstral coefficients with closely positioned (C-type) and separately located (S-type) electrode arrangements. Surface electrodes were bilaterally located on and between the sternocleidomastoid and the upper trapezius muscles. The EMG signals obtained during 20 repetitions of 10 motions were analyzed for each subject. Results from nine subjects showed that the mean recognition rate of the cepstral coefficients was at least 5% better than that of the AR coefficients for the S-type signals, while the improvement was less obvious for the C-type signals. The MMLM obtained the best discrimination results of the three discrimination methods. The S-type signals achieved higher recognition rates than the C-type signals in most cases. Among the various combinations of feature sets, classifiers and electrode arrangements proposed in this study, the combination of the cepstral coefficients and the MMLM with the S-type arrangement achieved the best discrimination efficiency. The proper choice of five of 10 motions raised the recognition rate to more than 97%.

The application of cepstral coefficients and maximum likelihood method in EMG pattern recognition.

A new technique for classifying patterns of movement via electromyographic (EMG) signals is presented. Two methods (conventional autoregressive (AR) coefficients and cepstral coefficients) for extracting features from EMG signals and three classification algorithms (Euclidean Distance Measure (EDM), Weighted Distance Measure (WDM), and Maximum Likelihood Method (MLM)) for discriminating signals representative of broad classes of movements are described and compared. These three classifiers are derived from Bayes classifier with some assumptions, the relationship among them is discussed. The conventional MLM is modified to avoid heavy matrix inversion. Six able-bodied subjects with two pairs of surface electrodes located on bilateral sternocleidomastoid and upper trapezius muscles were studied in the experiment. The EMG signals of 20 repetitions of 10 motions were analyzed for each subject. Experimental results showed that mean recognition rate of the cepstral coefficients was at least 5% superior to that of the AR coefficients. The improvement achieved by the cepstral method was statistically significant for all the three classifiers. Reasons for the superiority of cepstral features were investigated from the feature space and frequency domain, respectively. The cepstral coefficients owned better cluster separability in feature space and they emphasized the more informative part in the frequency domain. The discrimination rate of the MLM was the highest among three classifiers. Incorporation of the cepstral features with the MLM could reduce the misclassification rate by 10.6% when compared with the combination of AR coefficients and EDM. Proper choice of five of ten motions could further raise the recognition rate to more than 95%.