Python

Give me a functional python code that can do all of this, the audio files are $.$wav they are stored in a main directory called data and in this directory there are other directories containing the audio files for each digit. For example in $/$Data$/$Five all the files are of audio recordings of people saying five, so there are $9$ directories inside of Data:

The python code must do all of this:

Use feature extraction software to obtain the following from each audio, at least:

Zero crossing rate.

Spectral centroid.

Spectral rolloff.

MFCC coefficients.

Document these characteristics and obtain statistics $($mean$,$ variance, box plots$)$ for each one for the audios recorded in the laboratory.

Normalize the characteristics. Document the normalization process, its rationale, and the equation involved.

From the available data, separate $10\%$ as a test set.

Establish the centroid of the characteristics for each of the audio groups obtained $($excluding the test set$).$

For the elements of the test set, establish their distance to the centroid of each of the known groups. Set the prediction for each element as the smallest distance from it to the centroid of the groups.