Spaces:
Sleeping
Sleeping
File size: 2,402 Bytes
08c7f43 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 | import librosa
import numpy as np
import warnings
# Ignore librosa warnings about reading standard wav files to keep console clean
warnings.filterwarnings('ignore')
def preprocess_audio(file_path, target_sr, chunk_duration, step_duration, trim_top_db=20):
"""
Reads a raw audio file, cleans it, and splits it into fixed-size overlapping chunks.
Parameters:
- file_path: Path to the .wav file.
- target_sr: Target sample rate (16000Hz is standard for DL and fast).
- chunk_duration: Length of each slice in seconds (e.g., 3.0).
- step_duration: How far to slide the window in seconds (1.5s means 50% overlap).
Returns:
- np.array of shape (num_chunks, chunk_length)
"""
# STEP 1: Load & Resample (Handles Microphone Variations)
y, sr = librosa.load(file_path, sr=target_sr, res_type='soxr_hq')
# STEP 2: Silence Trimming (Handles Dead Space)
y_trimmed, _ = librosa.effects.trim(y, top_db=trim_top_db)
# STEP 3: Peak Normalization (Handles Volume Variations)
max_amplitude = np.max(np.abs(y_trimmed))
if max_amplitude > 0:
y_normalized = y_trimmed / max_amplitude
else:
y_normalized = y_trimmed
# STEP 4: Windowing / Slicing (Solution 2 - Prevents Data Loss)
chunk_length = int(target_sr * chunk_duration)
step_length = int(target_sr * step_duration)
chunks = []
# CASE A: The audio is shorter than our target chunk (e.g., 2 seconds)
if len(y_normalized) < chunk_length:
# Pad with zeros at the end to force it to be exactly 3 seconds
pad_length = chunk_length - len(y_normalized)
y_padded = np.pad(y_normalized, (0, pad_length), mode='constant')
chunks.append(y_padded)
# CASE B: The audio is longer (e.g., 7 seconds) -> Slice it!
else:
start = 0
while start + chunk_length <= len(y_normalized):
chunks.append(y_normalized[start : start + chunk_length])
start += step_length
# Catch the "tail" of the audio if it doesn't divide perfectly
# We take the LAST 3 seconds of the file to ensure we don't miss the ending
if start < len(y_normalized):
chunks.append(y_normalized[-chunk_length:])
# Return as a 2D numpy array: shape = (number_of_chunks, samples_per_chunk)
return np.array(chunks) |