Create Your First Project

Start adding your projects to your portfolio. Click on "Manage Projects" to get started

.pro-gallery-wix-wrapper {display: block !important;} .pro-gallery-wix-wrapper .gallery-item-container {opacity: 1 !important; display: block !important;}

Open Channel Sensing and Impulse Noise Filtering in Noisy Communication Systems

Project type

Digital Signal Processing

Date

Spring 2026

Developed a MATLAB-based communication systems simulation evaluating open-channel detection and impulse-noise filtering under varying noise and interference conditions. Compared MSE, adaptive thresholding, Welch PSD, and autocorrelation for channel detection using ROC/AUC accuracy and execution time, and evaluated threshold clipping, Hampel median, wavelet, and adaptive LMS filtering for corrupted speech using SNR and computational performance. Analyzed results across varying noise levels, channel counts, impulse amplitudes, and durations to identify performance tradeoffs. See the repository for a detailed methodology and results.

Area Under Curve performance surfaces. Click on this image for a link to the project repository.

This heatmap shows the detection accuracy (AUC) of each channel-sensing method across different background noise levels and numbers of channels. Higher AUC values indicate more accurate separation of occupied and open channels. The results show that detection accuracy remains high at low noise levels but decreases as noise increases, with Welch PSD generally providing the strongest performance in the intermediate noise range.

Detection accuracy vs noise across channels.

See how detection methods all perform relatively well at low noise levels, and poorly at high noise levels. As stated before, Welch PSD performs better at intermediate levels.

Computational Cost v Channel Count.

While Welch PSD and autocorrelation methods were seen to be more reliable at intermediate levels of noise, see how they are computationally more expensive. If given a system with minimal noise, it then suffices to choose a less demanding method such as mean squared error.

Filter effectiveness vs impulse noise width.

This figure compares the performance of several impulse-noise filtering methods applied to corrupted speech signals. It shows how threshold clipping, Hampel median, wavelet denoising, and adaptive LMS filtering reduce impulsive noise. The results highlight the differences in signal quality and filtering effectiveness, with LMS and Hampel filtering generally providing the strongest SNR improvement under the tested conditions.

.pro-gallery-wix-wrapper {display: block !important;} .pro-gallery-wix-wrapper .gallery-item-container {opacity: 1 !important; display: block !important;}

Back To Projects

Create Your First Project

Open Channel Sensing and Impulse Noise Filtering in Noisy Communication Systems

Project type

Date

Engineering Portfolio