dimension reduction of features

[bovitch]

I want to reduce the features dimension of speech signals. the used signals have different sizes.
how to get the features with fixed dimension after the dimensionality reduction??

 

[BradtheRad]

Perhaps you mean the amplitude of the waveforms?
Do you want to make loud words softer, and soft words louder?

Do you use a digital sound processing program? A lot of people like Audacity because it is free and has many of the features needed to make it useful.

The effect that might do what you want is called compression. It reduces the louder waveforms. You can set preferences which adjust the levels at which it operates.

 

[bovitch]

Thanks BradtheRad

I have performed a feature extraction for voice recognition
however the obtained features have different length (because the used signals have different length)
I want to reduced and fixed the length of the obtained features

 

[BradtheRad]

It sounds as though you want to change the time-length of each word. Is that correct?

Audacity can change speed without changing pitch, or vice-versa. It does this by re-sampling the data points.

You can perform the operation on an entire file, or just a portion.

I don't know of a way to apply it automatically so that you obtain fixed time lengths of words.