predict.py

import numpy as np
import os
import logging
import soundfile as sf
import librosa
import traceback

logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger(__name__)

def extract_features(file_path):
    """Extract audio features from a file."""
    try:
        logger.info(f"Starting feature extraction for: {file_path}")
        
        # Verify file exists
        if not os.path.exists(file_path):
            logger.error(f"File does not exist: {file_path}")
            return None
            
        # Verify file format
        try:
            with sf.SoundFile(file_path) as sf_file:
                logger.info(f"Audio file info: {sf_file.samplerate}Hz, {sf_file.channels} channels")
        except Exception as e:
            logger.error(f"Error reading audio file with soundfile: {str(e)}\n{traceback.format_exc()}")
            return None

        # Load audio file with error handling
        try:
            logger.info("Loading audio file...")
            y, sr = librosa.load(file_path, duration=30, sr=None)
            if len(y) == 0:
                logger.error("Audio file is empty")
                return None
            logger.info(f"Successfully loaded audio: {len(y)} samples, {sr}Hz sample rate")
        except Exception as e:
            logger.error(f"Error loading audio: {str(e)}\n{traceback.format_exc()}")
            return None

        # Ensure minimum duration
        duration = len(y) / sr
        logger.info(f"Audio duration: {duration:.2f} seconds")
        if duration < 1.0:
            logger.error("Audio file is too short (less than 1 second)")
            return None

        features_dict = {}
        
        try:
            # 1. MFCC (13 features x 2 = 26)
            logger.info("Extracting MFCC features...")
            mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13)
            features_dict['mfccs_mean'] = np.mean(mfccs, axis=1)
            features_dict['mfccs_var'] = np.var(mfccs, axis=1)
            logger.info(f"MFCC features shape: {mfccs.shape}")
        except Exception as e:
            logger.error(f"Error extracting MFCC: {str(e)}\n{traceback.format_exc()}")
            return None

        try:
            # 2. Chroma Features
            logger.info("Extracting chroma features...")
            chroma = librosa.feature.chroma_stft(y=y, sr=sr)
            features_dict['chroma'] = np.mean(chroma, axis=1)
            logger.info(f"Chroma features shape: {chroma.shape}")
        except Exception as e:
            logger.error(f"Error extracting chroma features: {str(e)}\n{traceback.format_exc()}")
            return None

        # Combine all features
        try:
            logger.info("Combining features...")
            features = np.concatenate([
                features_dict['mfccs_mean'],
                features_dict['mfccs_var'],
                features_dict['chroma']
            ])
            logger.info(f"Final feature vector shape: {features.shape}")
            return features
        except Exception as e:
            logger.error(f"Error combining features: {str(e)}\n{traceback.format_exc()}")
            return None

    except Exception as e:
        logger.error(f"Unexpected error in feature extraction: {str(e)}\n{traceback.format_exc()}")
        return None