File size: 14,110 Bytes
3f792e8 |
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 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 |
"""
Audio Preprocessing Module for Multilingual Audio Intelligence System
This module handles the standardization of diverse audio inputs into a consistent
format suitable for downstream ML models. It supports various audio formats
(wav, mp3, ogg, flac), sample rates (8k-48k), bit depths (4-32 bits), and
handles SNR variations as specified in PS-6 requirements.
Key Features:
- Format conversion and standardization
- Intelligent resampling to 16kHz
- Stereo to mono conversion
- Volume normalization for SNR robustness
- Memory-efficient processing
- Robust error handling
Dependencies: pydub, librosa, numpy
System Dependencies: ffmpeg (for format conversion)
"""
import os
import logging
import numpy as np
import librosa
from pydub import AudioSegment
from pydub.utils import which
from typing import Tuple, Optional, Union
import tempfile
import warnings
# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
# Suppress librosa warnings for cleaner output
warnings.filterwarnings("ignore", category=UserWarning, module="librosa")
class AudioProcessor:
"""
Handles audio preprocessing for the multilingual audio intelligence system.
This class standardizes diverse audio inputs into a consistent format:
- 16kHz sample rate (optimal for ASR models)
- Single channel (mono)
- Float32 numpy array format
- Normalized amplitude
"""
def __init__(self, target_sample_rate: int = 16000):
"""
Initialize AudioProcessor with target specifications.
Args:
target_sample_rate (int): Target sample rate in Hz. Default 16kHz
optimized for Whisper and pyannote models.
"""
self.target_sample_rate = target_sample_rate
self.supported_formats = ['.wav', '.mp3', '.ogg', '.flac', '.m4a', '.aac']
# Verify ffmpeg availability
if not which("ffmpeg"):
logger.warning("ffmpeg not found. Some format conversions may fail.")
def process_audio(self, audio_input: Union[str, bytes, np.ndarray],
input_sample_rate: Optional[int] = None) -> Tuple[np.ndarray, int]:
"""
Main processing function that standardizes any audio input.
Args:
audio_input: Can be file path (str), audio bytes, or numpy array
input_sample_rate: Required if audio_input is numpy array
Returns:
Tuple[np.ndarray, int]: (processed_audio_array, sample_rate)
Raises:
ValueError: If input format is unsupported or invalid
FileNotFoundError: If audio file doesn't exist
Exception: For processing errors
"""
try:
# Determine input type and load audio
if isinstance(audio_input, str):
# File path input
audio_array, original_sr = self._load_from_file(audio_input)
elif isinstance(audio_input, bytes):
# Bytes input (e.g., from uploaded file)
audio_array, original_sr = self._load_from_bytes(audio_input)
elif isinstance(audio_input, np.ndarray):
# Numpy array input
if input_sample_rate is None:
raise ValueError("input_sample_rate must be provided for numpy array input")
audio_array = audio_input.astype(np.float32)
original_sr = input_sample_rate
else:
raise ValueError(f"Unsupported input type: {type(audio_input)}")
logger.info(f"Loaded audio: {audio_array.shape}, {original_sr}Hz")
# Apply preprocessing pipeline
processed_audio = self._preprocess_pipeline(audio_array, original_sr)
logger.info(f"Processed audio: {processed_audio.shape}, {self.target_sample_rate}Hz")
return processed_audio, self.target_sample_rate
except Exception as e:
logger.error(f"Audio processing failed: {str(e)}")
raise
def _load_from_file(self, file_path: str) -> Tuple[np.ndarray, int]:
"""Load audio from file path."""
if not os.path.exists(file_path):
raise FileNotFoundError(f"Audio file not found: {file_path}")
file_ext = os.path.splitext(file_path)[1].lower()
if file_ext not in self.supported_formats:
raise ValueError(f"Unsupported format {file_ext}. Supported: {self.supported_formats}")
try:
# Use librosa for robust loading with automatic resampling
audio_array, sample_rate = librosa.load(file_path, sr=None, mono=False)
return audio_array, sample_rate
except Exception as e:
# Fallback to pydub for format conversion
logger.warning(f"librosa failed, trying pydub: {e}")
return self._load_with_pydub(file_path)
def _load_from_bytes(self, audio_bytes: bytes) -> Tuple[np.ndarray, int]:
"""Load audio from bytes (e.g., uploaded file)."""
# Create temporary file for processing
with tempfile.NamedTemporaryFile(delete=False, suffix='.audio') as tmp_file:
tmp_file.write(audio_bytes)
tmp_path = tmp_file.name
try:
# Try to detect format and load
audio_array, sample_rate = self._load_with_pydub(tmp_path)
return audio_array, sample_rate
finally:
# Clean up temporary file
try:
os.unlink(tmp_path)
except OSError:
pass
def _load_with_pydub(self, file_path: str) -> Tuple[np.ndarray, int]:
"""Load audio using pydub with format detection."""
try:
# Let pydub auto-detect format
audio_segment = AudioSegment.from_file(file_path)
# Convert to numpy array
samples = np.array(audio_segment.get_array_of_samples(), dtype=np.float32)
# Handle stereo audio
if audio_segment.channels == 2:
samples = samples.reshape((-1, 2))
# Normalize to [-1, 1] range
samples = samples / (2**15) # 16-bit normalization
return samples, audio_segment.frame_rate
except Exception as e:
raise Exception(f"Failed to load audio with pydub: {str(e)}")
def _preprocess_pipeline(self, audio_array: np.ndarray, original_sr: int) -> np.ndarray:
"""
Apply the complete preprocessing pipeline.
Pipeline steps:
1. Convert stereo to mono
2. Resample to target sample rate
3. Normalize amplitude
4. Apply basic noise reduction (optional)
"""
# Step 1: Convert to mono if stereo
if len(audio_array.shape) > 1 and audio_array.shape[0] == 2:
# librosa format: (channels, samples) for stereo
audio_array = np.mean(audio_array, axis=0)
elif len(audio_array.shape) > 1 and audio_array.shape[1] == 2:
# pydub format: (samples, channels) for stereo
audio_array = np.mean(audio_array, axis=1)
# Ensure 1D array
audio_array = audio_array.flatten()
logger.debug(f"After mono conversion: {audio_array.shape}")
# Step 2: Resample if necessary
if original_sr != self.target_sample_rate:
audio_array = librosa.resample(
audio_array,
orig_sr=original_sr,
target_sr=self.target_sample_rate,
res_type='kaiser_best' # High quality resampling
)
logger.debug(f"Resampled from {original_sr}Hz to {self.target_sample_rate}Hz")
# Step 3: Amplitude normalization
audio_array = self._normalize_audio(audio_array)
# Step 4: Basic preprocessing for robustness
audio_array = self._apply_preprocessing_filters(audio_array)
return audio_array.astype(np.float32)
def _normalize_audio(self, audio_array: np.ndarray) -> np.ndarray:
"""
Normalize audio amplitude to handle varying SNR conditions.
Uses RMS-based normalization for better handling of varying
signal-to-noise ratios (-5dB to 20dB as per PS-6 requirements).
"""
# Calculate RMS (Root Mean Square)
rms = np.sqrt(np.mean(audio_array**2))
if rms > 0:
# Target RMS level (prevents over-amplification)
target_rms = 0.1
normalization_factor = target_rms / rms
# Apply normalization with clipping protection
normalized = audio_array * normalization_factor
normalized = np.clip(normalized, -1.0, 1.0)
logger.debug(f"RMS normalization: {rms:.4f} -> {target_rms:.4f}")
return normalized
return audio_array
def _apply_preprocessing_filters(self, audio_array: np.ndarray) -> np.ndarray:
"""
Apply basic preprocessing filters for improved robustness.
Includes:
- DC offset removal
- Light high-pass filtering (removes very low frequencies)
"""
# Remove DC offset
audio_array = audio_array - np.mean(audio_array)
# Simple high-pass filter to remove very low frequencies (< 80Hz)
# This helps with handling background noise and rumble
try:
from scipy.signal import butter, filtfilt
# Design high-pass filter
nyquist = self.target_sample_rate / 2
cutoff = 80 / nyquist # 80Hz cutoff
if cutoff < 1.0: # Valid frequency range
b, a = butter(N=1, Wn=cutoff, btype='high')
audio_array = filtfilt(b, a, audio_array)
logger.debug("Applied high-pass filter (80Hz cutoff)")
except ImportError:
logger.debug("scipy not available, skipping high-pass filter")
except Exception as e:
logger.debug(f"High-pass filter failed: {e}")
return audio_array
def get_audio_info(self, audio_input: Union[str, bytes]) -> dict:
"""
Get detailed information about audio file without full processing.
Returns:
dict: Audio metadata including duration, sample rate, channels, etc.
"""
try:
if isinstance(audio_input, str):
# File path
if not os.path.exists(audio_input):
raise FileNotFoundError(f"Audio file not found: {audio_input}")
audio_segment = AudioSegment.from_file(audio_input)
else:
# Bytes input
with tempfile.NamedTemporaryFile(delete=False) as tmp_file:
tmp_file.write(audio_input)
tmp_path = tmp_file.name
try:
audio_segment = AudioSegment.from_file(tmp_path)
finally:
try:
os.unlink(tmp_path)
except OSError:
pass
return {
'duration_seconds': len(audio_segment) / 1000.0,
'sample_rate': audio_segment.frame_rate,
'channels': audio_segment.channels,
'sample_width': audio_segment.sample_width,
'frame_count': audio_segment.frame_count(),
'max_possible_amplitude': audio_segment.max_possible_amplitude
}
except Exception as e:
logger.error(f"Failed to get audio info: {e}")
return {}
# Utility functions for common audio operations
def validate_audio_file(file_path: str) -> bool:
"""
Quick validation of audio file without full loading.
Args:
file_path (str): Path to audio file
Returns:
bool: True if file appears to be valid audio
"""
try:
processor = AudioProcessor()
info = processor.get_audio_info(file_path)
return info.get('duration_seconds', 0) > 0
except Exception:
return False
def estimate_processing_time(file_path: str) -> float:
"""
Estimate processing time based on audio duration.
Args:
file_path (str): Path to audio file
Returns:
float: Estimated processing time in seconds
"""
try:
processor = AudioProcessor()
info = processor.get_audio_info(file_path)
duration = info.get('duration_seconds', 0)
# Rough estimate: 0.1x to 0.3x real-time for preprocessing
# depending on format conversion needs
estimated_time = duration * 0.2
return max(estimated_time, 1.0) # Minimum 1 second
except Exception:
return 10.0 # Default estimate
if __name__ == "__main__":
# Example usage and testing
processor = AudioProcessor()
# Test with a sample file (if available)
test_files = ["sample.wav", "sample.mp3", "test_audio.flac"]
for test_file in test_files:
if os.path.exists(test_file):
try:
print(f"\nTesting {test_file}:")
# Get info
info = processor.get_audio_info(test_file)
print(f"Info: {info}")
# Process
audio, sr = processor.process_audio(test_file)
print(f"Processed: shape={audio.shape}, sr={sr}")
# Validate
is_valid = validate_audio_file(test_file)
print(f"Valid: {is_valid}")
except Exception as e:
print(f"Error processing {test_file}: {e}") |