Librosa

Librosa: The essential Python package for music and audio analysis, providing core building blocks for Music Information Retrieval (MIR) systems.

Librosa is your high-level Python library for serious audio work. It handles the heavy lifting: load files (WAV, MP3, FLAC) into NumPy arrays, typically resampling to a 22050 Hz default, then execute complex signal processing. We use it to extract critical features like Mel-Frequency Cepstral Coefficients (MFCCs), chromagrams, and Mel spectrograms. Key functionality includes tempo estimation, beat tracking, and onset detection. For advanced tasks, leverage its Harmonic-Percussive Source Separation (HPSS) to isolate tonal and transient components, a capability crucial for robust MIR and speech recognition applications.

https://librosa.org/