BF-Streams: Native stream for audio data

Notes after brainstorming:

After considering multiple libraries, Librosa is the most commonly used in Python and has most of the important features. Librosa is also reusable. Hence, we can integrate Librosa into MLPro for loading and processing audio data, rather than reinventing the wheel
There are 3 main components of audio data: amplitude, time, and frequency
We can load the audio data through librosa.load(....) from .wav format, which returns samples and sampling rate. For mp3 format, it can be done using a converter (e.g. from pydub import AudioSegment)
librosa has several types of visualizations (librosa.display.[plot type]), which can be incorporated into our MLPro-Streams visualization
Fourier transform -> converts a continuous signal from time-domain (x-axis = time, y-axis = amplitude) to frequency-domain (x-axis = frequency, y-axis = magnitude)
Spectrogram (x-axis = time, y-axis = frequency, z-axis (colour) = amplitude)-> can be generated using Short-Time Fourier Transform (STFT). This is also available in librosa. Spectrogram shows the signal strength, or “loudness”, of a signal over time at various frequencies present in a particular waveform (https://pnsn.org/spectrograms/what-is-a-spectrogram#:~:text=A%20spectrogram%20is%20a%20visual,energy%20levels%20vary%20over%20time.).
Interesting features in librosa:
- data trimming and zooming
- spectral feature : chroma, Melspectra, mfcc, tonnetz, etc (librosa.feature.[name of the feature])
- splitting original data into harmonic and percussive (librosa.effects.hpss(...))
- audio data analysis (beat)
- many more
Real-time audio data processing from a microphone requires an extra package (candidate: https://pypi.org/project/PyAudio/)

fhswf / MLPro

BF-Streams: Native stream for audio data #562