Accurate Calculation of One-Third Octave Bands for Speech and Noise Signals
Precise computation of one-third octave bands for speech or noise signals with implementation insights
LMS Algorithm Implementation for Filtering
Implementation of LMS algorithm for filtering with weight vector reconstruction, enabling noise signal reconstruction and time delay calculation through adaptive filtering techniques
MATLAB Implementation of Blind Source Separation Algorithm
Blind Source Separation Algorithm capable of extracting original signals from mixed signals. Verified with MATLAB: The program successfully separates sinusoidal signals and random noise signals from their mixture. The implementation ensures correct execution through comprehensive testing and includes signal generation, mixing matrix configuration, and separation performance evaluation.
Simulation of Radar Noise Signals
This work explores the simulation of radar noise signals, sharing key methodologies and practical implementation approaches through code examples and algorithm explanations.
Wavelet Transform and Energy Distribution Characteristics of Noisy Signals for Image Denoising
Based on the energy distribution characteristics of wavelet transform and noisy signals, we propose a method that first decomposes noisy images using multi-scale wavelet transform, calculates the noise variance and thresholds for high-frequency coefficients at each scale, processes the high-frequency coefficients using these scale-specific thresholds, and then reconstructs the image using wavelet coefficients to achieve effective image denoising.
Frequency Estimation of Sinusoidal Signals in Noise using Pisarenko Harmonic Decomposition, MUSIC Algorithm, and ESPRIT Algorithm
This project implements frequency estimation of sinusoidal signals embedded in Gaussian white noise through three high-resolution spectral estimation methods: Pisarenko Harmonic Decomposition, MUSIC (Multiple Signal Classification) algorithm, and ESPRIT (Estimation of Signal Parameters via Rotational Invariance Techniques) algorithm. The sinusoidal signal is defined with specific frequency components, while the additive white Gaussian noise has controlled variance. Using 128 data samples, the implementation involves: 1) Performing 20 independent trials with each algorithm to record frequency estimates and compute statistical mean and variance; 2) Analyzing algorithm performance under increasing noise power conditions to evaluate robustness and accuracy.