Extract-Audio-Feature---Python (PAWAN AGRAHARI)

Overview: This project extracts various audio features (like Energy, Power, MFCCs, and Spectral Features) from audio files using Python. It also provides functionality to process video files by extracting audio and subsequently analyzing it for audio features.

Dependencies: The project uses the following Python libraries:

Package Description:

moviepy - Used to process video files and extract audio from them.

librosa - A library for analyzing and extracting audio features.

numpy - Numerical computing library for efficient data manipulation.

scipy - Scientific computing tools used for mathematical operations.

soundfile - Handles reading and writing of audio files.

matplotlib - Used to visualize audio waveforms and extracted features.

Methodology

1. Audio and Video Processing If the input is a video file, the script uses moviepy to extract audio from the video. Audio is saved temporarily as a .wav file for further analysis.

from moviepy.editor import VideoFileClip

video_path = "sample_videos/video.mp4" output_audio_path = "output/audio_from_video.wav"

clip = VideoFileClip(video_path) clip.audio.write_audiofile(output_audio_path)

2. Audio Feature Extraction Using librosa, the script extracts key features:

Energy and Power Mel Frequency Cepstral Coefficients (MFCCs) Spectral Features (Spectral Centroid, Bandwidth, and Roll-off)

Code for Extracting Features import librosa import numpy as np

Load audio file

audio_path = "sample_audio/audio_file.wav" y, sr = librosa.load(audio_path)

Energy and Power

energy = np.sum(np.square(y)) print(f"Energy: {energy}")

Extract MFCCs

mfccs = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=13) print(f"MFCCs Shape: {mfccs.shape}")

Spectral Features

centroid = librosa.feature.spectral_centroid(y=y, sr=sr) bandwidth = librosa.feature.spectral_bandwidth(y=y, sr=sr) rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr)

3. Visualization The script uses matplotlib to plot:

The audio waveform. The extracted MFCCs as a heatmap.

Code for Visualization import matplotlib.pyplot as plt

Plot audio waveform

plt.figure(figsize=(10, 4)) plt.plot(y) plt.title("Audio Waveform") plt.xlabel("Samples") plt.ylabel("Amplitude") plt.show()

Plot MFCCs

plt.figure(figsize=(10, 4)) librosa.display.specshow(mfccs, sr=sr, x_axis='time') plt.colorbar() plt.title("MFCCs") plt.show()

Conclusion:

This project demonstrates how to extract audio features and visualize them using Python. It uses libraries like moviepy for video processing and librosa for audio analysis, making it ideal for audio-related applications like speech recognition, music analysis, or feature engineering for machine learning.