How to arrange data for time series classification learning

I have several EMG recordings from test subjects (voltage as a function of time). That is, the data is time series data (see below).

Each recording is e.g. 27.000 samples long and correspond to a certain exercise which the subject performed. Each exercise has been recorded several times. So the exercise 'abduction' is e.g. a 27.000 x 10 matrix (10 repetitions of 27.000 samples). This is the case for all exercises.

I wanted to use the Matlab's classification learner app, but i struggle with how to arrange my data, so that the app will read it as: the response is the type of exercise (e.g. 'abduction') and the predictor is the 27.000 samples for each repetitions of the exercise.

Or am I going about this all wrong? As I understood it, each time instant is a variable, e.g. millisecond 1, millisecond 2, millisecond 3 etc. is a variable, yielding 27.000 variables. Should i use another application in matlab?

Answer 1

[Prashant Arora]

Hi Malte,

I believe you will need to pre-process the time series data to extract features for the classification learner. Some examples of features from time-series can be:

Minimum/Maximum, Means, Standard Deviations, Cross-Correlations etc.