Audio Processing

Introduction

Audio processing analyzes and transforms audio signals using libraries like librosa for feature extraction.

Loading Audio

import librosa
import numpy as np

# Load audio file
y, sr = librosa.load('audio.wav')
print(f"Duration: {len(y)/sr:.2f}s, Sample rate: {sr}")

# Load with specific duration
y, sr = librosa.load('audio.wav', sr=22050, offset=0.0, duration=10)

# Generate tone
sr = 22050
tone = librosa.tone(440, sr=sr, duration=1)

Audio Features

import librosa

# Spectrogram
S = librosa.feature.melspectrogram(y=y, sr=sr)
S_dB = librosa.power_to_db(S, ref=np.max)

# MFCC
mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13)

# Chroma
chroma = librosa.feature.chroma_stft(y=y, sr=sr)

# Spectral features
spectral_centroid = librosa.feature.spectral_centroid(y=y, sr=sr)
spectral_rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr)

# Zero crossing rate
zcr = librosa.feature.zero_crossing_rate(y)

Audio Effects

import librosa.effects

# Trim silence
y_trimmed, idx = librosa.effects.trim(y, top_db=20)

# Time stretch
y_stretched = librosa.effects.time_stretch(y, rate=1.5)

# Pitch shift
y_shifted = librosa.effects.pitch_shift(y, sr, n_steps=2)

Audio Synthesis

import numpy as np

# Generate audio
sr = 22050
t = np.linspace(0, 1, sr)
tone = np.sin(2 * np.pi * 440 * t)

# Save audio
import soundfile as sf
sf.write('tone.wav', tone, sr)

Practice Problems

1. Load audio files
2. Extract MFCC features
3. Compute spectrogram
4. Apply time stretching
5. Save audio to file