Hand gesture recognition using smooth wavelet packet transformation and hybrid CNN based on surface EMG and accelerometer signal

Surface electromyography (sEMG), which has the advantages of being simple to acquire and quick to respond, is frequently utilized in domains like human–computer interface and prosthetic control as a control source for gesture recognition. Firstly, we propose a method to decompose the sEMG into the time–frequency domain information using the smooth wavelet packet transform (SWPT), which has a faster processing speed compared to previous methods, such as the continuous wavelet transform (CWT) and wavelet packet transform (WPT), requiring only 12% of the time consumption of CWT and 66% of WPT. Secondly, to increase the recognition accuracy of hand gestures, a network model was built using a combination of convolutional neural network (CNN), long short term memory (LSTM), and convolutional block attention module (CBAM) with the accelerometer (ACC) data fusion. With an average accuracy of 92.159%, this approach significantly outperformed other similar research studies when evaluated on the public dataset NapiroDB5.