marc nicole <
mk1853387@gmail.com> wrote or quoted:
I have a hard time finding a way to check whether audio data samples are >containing empty noise or actual significant voice/noise.
Or, you could have a human do a quick listen to some audio files to
gauge the "empty-noise ratio," then use that number as the filename
as a float, and finally train up a neural net on this. E.g.,
0.99.wav # very empty
0.992.wav # very empty file #2
0.993.wav # very empty file #3
0.00.wav # very not empty file
0.002.wav # very not empty file #2
One possible approach:
import os
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.optimizers import Adam
import librosa
## Data Preparation
# Function to extract audio features
def extract_features(file_path):
audio, sr = librosa.load(file_path)
mfccs = librosa.feature.mfcc(y=audio, sr=sr, n_mfcc=13)
return np.mean(mfccs.T, axis=0)
# Load data from directory
directory = 'd' # for example
X = []
y = []
for filename in os.listdir(directory):
if filename.endswith('.wav'):
file_path = os.path.join(directory, filename)
X.append(extract_features(file_path))
y.append(float(filename[:-4])) # Assuming filename is the p value
X = np.array(X)
y = np.array(y)
# Split data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Feature scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
## Neural Network Model
model = Sequential([
Dense(64, activation='relu', input_shape=(13,)),
Dense(32, activation='relu'),
Dense(1)
])
model.compile(optimizer=Adam(learning_rate=0.001), loss='mean_squared_error')
## Training
model.fit(X_train_scaled, y_train, epochs=100, batch_size=32, validation_split=0.2, verbose=1)
## Evaluation
test_loss = model.evaluate(X_test_scaled, y_test, verbose=0)
print(f"Test Loss: {test_loss}")
## Prediction Function
def predict_p(audio_file):
features = extract_features(audio_file)
scaled_features = scaler.transform(features.reshape(1, -1))
prediction = model.predict(scaled_features)
return prediction[0][0]
# Example usage
new_audio_file = 'path/to/new/audio/file.wav'
predicted_p = predict_p(new_audio_file)
print(f"Predicted p value: {predicted_p}")
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)