Metrological Aspects of SEMG Signal Acquisition, Processing, and Application Design

Significant growth in SEMG-based applications has been observed over the last few decades. It is primarily due to advances in electronics fabrication and computational capabilities. However, the metrological issues and standardization issues are still untouched and need to be addressed critically. This chapter attempts to compile the recommendations, standards, and requirements of such type of systems, as well as a case study of SEMG operated lower limb prosthesis design. The SENIAM recommendations are thoroughly discussed and implemented in the application design. A high-level and low-level control module is used to implement a biologically inspired control structure. The high-level control module is responsible for estimating activity intention, whereas the low-level control module is responsible for estimating the desired angular position of the joints. Both the control modules utilize SEMG signal differently. The high-level control module problem is formulated as classification problem, whereas the low-level control module problem is implemented as classification problem for knee angle prediction and continuous value estimation problem for ankle angle estimation. SVM, LDA, and NN classifiers with time domain feature vector were found suitable for high-level control module. For knee angle prediction, SVMQ was found to be superior, whereas for ankle angle estimation, the autoregressive exogenous neural network (NARX) model with SEMG and knee angle information depicts satisfactory performance over the different locomotion modes.