Friday, May 22, 2020

In this blog we will discuss  about sEMG signal classification using CNN with
transfer learning.

We will see one by one clearly.


first what is sEMG ?

Electromyography is an electrodiagnostic medicine technique for evaluating and recording the electrical activity produced by skeletal muscles.

Simply saying, When muscles contraction an elactical plus will be created , we record them using 
elctrodes and emg sensors,



this will create non-stationary signals

we can claasify our sugnals into two types,

1. stationary signals 
2.non-stationary signals

stationary singals means with the time the features of the signals never change
example amplitde and frequency 


but non stationry signals will change over time ,



there are two domains in signals 
1.time domain
2.frequency domain




time domain plot signals time(x_axis) versus amplitude (y_axis) or strength(y_axis)
we can gether information about amplitude

problems is we cant decompose a noicy singal,





frequency domain plots frequency (x_axis) vs amplitude(y_axis)

so we can easily decompose signals according to their frequency,

example



we will use Fourier transform to convert time domain to frequency domain




we are going to use a freely available data set for this problem