Update feature_extractor.py

[BBC-Esq]

Added mel filter bank caching to FeatureExtractor class to optimize memory usage and reduce computational overhead when processing multiple audio files with identical parameters, particularly beneficial for batch processing scenarios.

[abodacs]

@BBC-Esq
Thanks for your work

[abodacs]

Why was the hann_window deleted?

[BBC-Esq]

I'll take a look...

[BBC-Esq]

This is actually an optimization. In the old version, a new Hann window was being created and moved to the device every time call was executed. The new version creates it once during initialization and caches it as an instance variable (self.window).

[abodacs]

Improved readability, thank you.

[MahmoudAshraf97]

Hello, the way you implemented caching is not going to work because the only parameter that might change when creating the mel filters is n_mels, to change this parameter you need to initialize a new FeatureExtractor instance which will clear cache thus invalidating its purpose. Anyways, creating window takes around 6us which represents less than 0.05% of the execution time and the mel filters is created only once thus there is no need for caching.

In the future, it's preferable to use a caching decorator such as functools.lru_cache as an easy caching solution instead of implementing it as a dictionary