Ingestive Pattern Recognition on Cattle Using EMG Segmentation and Feature Extraction

This work presents a non-invasive method to identify the ingestive behavior in ruminants using Surface Electromyography (sEMG) of the masseter muscle. It was evaluate whether the rumination and intake process could be distinguished using sEMG signal features and machine learning techniques. To collect chewing sEMG signal, superficial Ag/AgCl electrodes were placed on two ruminant animals masseter muscle and the data was sampled during eating with an analog-to-digital converter. Three segmentation techniques was explored and applied to automatically subdivide the chewing movement signal. Five classifiers (k-nn, LDA, SVM, NB and DT) were evaluated using four features (RMS, SSC, ZC and WL) extracted from the signal. We found and accuracy over 93% using a fixed length segmentation method and k-nn for classification with k > 7. Future works may explores the implementation of time series analysis to improve the classifier performance.