Gestures made by a human can be classified using Electromyography (EMG) sig-nals collected from the forearm; even with low frequency devices. Numerous steps are required from data collection and pre-processing through to final classi-fication. Traditionally, an important part of EMG signal classification is extract-ing features from the raw signal to reduce dimensionality. It is predominantly car-ried out manually before the signals are input into a neural network. In this re-search we successfully used a CNN to extract the features automatically and an LSTM layer was utilized to classify the gestures. This network architecture re-moves a step in the gesture classification process. Using the raw signals input into a CNN/LSTM hybrid increased classification when compared with an LSTM net-work that required features to be manually extracted from the raw signals.
|Title of host publication||Proceedings of the International Conference on BIg Data, IoT and Machine Learning|
|Editors||Mohammad Shamsul Arefin, M Shamim Kaiser, Anirban Bandyopadhyay, Md Atiqur Rahman Ahad, Kanad Ray|
|Number of pages||18|
|Publication status||Published - 4 Dec 2021|
|Name||Lecture Notes on Data Engineering and Communications Technologies|