test_eval_speaker_identification.py

import os
import numpy as np
import librosa
from speaker.speaker_identification import assign_speaker_for_audio_list

# Define constants
TEST_DATA_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'Test_data_for_clas_Idef')
AUDIO_FILES_DIR = os.path.join(TEST_DATA_DIR, 'enni_audio_files')
NUMPY_FILES_DIR = os.path.join(TEST_DATA_DIR, 'enni_testset_numpy_minimal')
FILEPATHS_DIR = os.path.join(TEST_DATA_DIR, 'enni_testset_filepaths_minimal')

def generate_fake_audio_test_set(num_samples=10, length=16000, random_seed=42):
    """
    Generate a synthetic test set of fake audio signals (numpy arrays).
    Args:
        num_samples (int): Number of audio samples.
        length (int): Length of each audio sample (e.g., 1 second at 16kHz).
        random_seed (int): Seed for reproducibility.
    Returns:
        List[np.ndarray]: List of fake audio signals.
    """
    np.random.seed(random_seed)
    return [np.random.randn(length) for _ in range(num_samples)]

def test_file_paths():
    """Test with all real audio files from the dataset"""
    # Get file paths using the constant
    audio_dir = AUDIO_FILES_DIR
    
    # Get all child and adult files
    child_files = [
        os.path.join(audio_dir, file) 
        for file in os.listdir(audio_dir) 
        if file.startswith('child_') and file.endswith('.wav')
    ]  # Use all child files
    
    adult_files = [
        os.path.join(audio_dir, file) 
        for file in os.listdir(audio_dir) 
        if file.startswith('adult_') and file.endswith('.wav')
    ]  # Use all adult files
    
    # Create list with known order
    audio_list = child_files + adult_files
    
    # Get speaker IDs
    speaker_ids = assign_speaker_for_audio_list(audio_list)
    
    # Print results
    print("\n--- Testing with file paths ---")
    print(f"Testing {len(audio_list)} audio files: {len(child_files)} child files and {len(adult_files)} adult files")
    
    # Count correct predictions
    correct = 0
    for i, (file, speaker_id) in enumerate(zip(audio_list, speaker_ids)):
        expected = "Speaker_id_0" if "child_" in file else "Speaker_id_1"
        is_correct = speaker_id == expected
        correct += 1 if is_correct else 0
        
        # Print only the first 5 examples to avoid cluttering the output
        if i < 5:
            print(f"{i+1}. {os.path.basename(file)}: {speaker_id} (Expected: {expected}) {'✓' if is_correct else '✗'}")
    
    # Print accuracy
    accuracy = correct / len(audio_list) * 100 if audio_list else 0
    print(f"Accuracy: {correct}/{len(audio_list)} ({accuracy:.2f}%)")

def test_numpy_arrays():
    """Test with NumPy arrays by loading all audio files"""
    # Get file paths using the constant
    audio_dir = AUDIO_FILES_DIR
    
    # Load all child and adult files as arrays
    child_files = [
        os.path.join(audio_dir, file) 
        for file in os.listdir(audio_dir) 
        if file.startswith('child_') and file.endswith('.wav')
    ]
    
    adult_files = [
        os.path.join(audio_dir, file) 
        for file in os.listdir(audio_dir) 
        if file.startswith('adult_') and file.endswith('.wav')
    ]
    
    # Load as arrays
    child_arrays = [librosa.load(f, sr=16000)[0] for f in child_files]
    adult_arrays = [librosa.load(f, sr=16000)[0] for f in adult_files]
    
    # Create list with known order
    audio_list = child_arrays + adult_arrays
    filenames = [os.path.basename(f) for f in child_files + adult_files]
    
    # Get speaker IDs
    speaker_ids = assign_speaker_for_audio_list(audio_list)
    
    # Print results
    print("\n--- Testing with NumPy arrays ---")
    print(f"Testing {len(audio_list)} audio arrays: {len(child_arrays)} child arrays and {len(adult_arrays)} adult arrays")
    
    # Count correct predictions
    correct = 0
    for i, (filename, speaker_id) in enumerate(zip(filenames, speaker_ids)):
        expected = "Speaker_id_0" if "child_" in filename else "Speaker_id_1"
        is_correct = speaker_id == expected
        correct += 1 if is_correct else 0
        
        # Print only the first 5 examples to avoid cluttering the output
        if i < 5:
            print(f"{i+1}. {filename} (as array): {speaker_id} (Expected: {expected}) {'✓' if is_correct else '✗'}")
    
    # Print accuracy
    accuracy = correct / len(audio_list) * 100 if audio_list else 0
    print(f"Accuracy: {correct}/{len(audio_list)} ({accuracy:.2f}%)")

if __name__ == "__main__":
    # Test with synthetic data
    print("--- Testing with synthetic data ---")
    audio_list = generate_fake_audio_test_set(num_samples=5)
    speaker_ids = assign_speaker_for_audio_list(audio_list)
    print(f"Synthetic data predictions: {speaker_ids}")
    
    # Test with real files
    try:
        test_file_paths()
    except Exception as e:
        print(f"Error testing file paths: {e}")
    
    # Test with NumPy arrays
    try:
        test_numpy_arrays()
    except Exception as e:
        print(f"Error testing NumPy arrays: {e}")