Cepstrum Computation for Speech Signals
Speech signal cepstrum computation program designed for enhanced extraction of pitch and formant information through advanced spectral analysis techniques
MATLAB-Based Speech Recognition System
This simple, compact MATLAB code implements speech recognition by comparing input audio signals with pre-stored voice templates; if a successful match occurs, an "access granted" signal is generated, otherwise an "access denied" signal is produced.
Short-Term Analysis of Speech Signals
Short-term analysis of speech signals includes key components such as frame splitting, short-term energy, short-term average magnitude, short-term zero-crossing rate, short-term autocorrelation function, short-term magnitude difference, cepstrum, complex cepstrum, LPC coefficients, and LPC spectral estimation. These fundamental programs assigned by my supervisor after securing postgraduate admission ensure absolute quality through robust implementation.
Extracting Speech Signal Spectral Envelope and Pitch Frequency Using Cepstrum Analysis
Implementing cepstral analysis to extract spectral envelope and pitch frequency from speech signals, with MATLAB code implementation insights.
Speech Signal Denoising Using Wavelet Transform
MATLAB-based implementation utilizing wavelet transform for speech signal denoising with practical code demonstration
Short-Time Analysis of Speech Signals
Speech signals are time-varying in nature, with individual parameter variations occurring more gradually than the signal itself. Consequently, measuring these parameters requires a significantly lower sampling frequency compared to the signal's original sampling rate. Through window function weighting, the signal is segmented in the time domain into local signal sequences for measurement. Proper short-time analysis requires defining two key dimensions: window length (duration of the weighted signal segment) and measurement interval (frame rate, representing the spacing between consecutive windows). Core short-time analysis operations include short-time energy (reflecting amplitude variations), short-time autocorrelation function (detecting periodicity), and short-time zero-crossing rate.
Blind Source Separation of Speech Signals in Real Environments
This is the latest program code for blind source separation of speech signals in real environments, implementing Independent Component Analysis (ICA) for speech signal processing. After extraction, simply run the program with recorded mixed speech signals as input to observe the separation results. The code employs advanced ICA algorithms to extract independent components from mixed audio sources.
Linear Predictive Coding of Speech Signals
Implementation of Linear Predictive Coding for speech signals, featuring a main program that orchestrates four subprograms handling prediction analysis, quantization, encoding, and decoding functionalities.
Independent Component Analysis (ICA) Algorithm
AMUSE - An Independent Component Analysis (ICA) Algorithm for Blind Separation of Mixed Speech Signals, implementing second-order statistics and time-delayed covariance matrices for source separation
Feature Classification of Speech Signals Using BP Neural Network
Implementation of effective classification for folk music, guzheng (Chinese zither), rock, and pop music genres through BP neural network architecture with feature extraction and pattern recognition techniques