Speech signal denoising

Project Title

Speech signal denoising

Project Area of Specialization Information & Communication TechnologyProject Summary

Speech denoising technique removes noise from speech signals while improving speech quality. This aim of this project is to shows how deep learning networks can be used to remove different types of noise from the speech signals.It is well understood that noisy environments degrade the quality and intelligibility of speech signals. Speech denoising entails reducing or eliminating the noisy part of the speech as well as distinguishing useful speech from the speech and noise mixture.

Project Objectives

This project expects to give the most ideal noises removal method for de-noising and recovering clean signal from an uproarious noisy signal. The point is to utilize diverse denoising procedures and compare their results to come to an end result with respect to which one of them is the most ideal for improving voice signals. The examination is done by assessing the performance of various denoising procedures for various kinds of speech samples. This assessment is finished by adding irregular noise to the speech signal then, at that point, applying denoising procedures to get denoised speech signal. A correspondence is drawn between unique signal and denoised signal through assessment boundaries like SNR and PSNR.

Project Implementation Method

The procedures which have been utilized are Weiner Filter, Wavelets denoising and Deep Learning techniques. The performance assessment will be done based on variation of evaluation parameters (SNR and PSNR values) for various denoising methods.

Benefits of the Project

Hearing comes second only to vision as a sensory process for obtaining critical information during different aircraft operations. The aviation environment has a variety of noise sources, both in the air and on the ground. Noise exposure became a critical problem for pilots. The findings of this project would demonstrate that the mentioned approaches can improve speech performance in the presence of significant aviation noise. In windy locations, an unique sort of noise emerges when the air stream creates a very non-stationary disruption in the recorded signal. The aforementioned strategies would be a well-established way for reducing stationary background noise signals. It would aid in the suppression of wind noise signals that change rapidly.

Technical Details of Final Deliverable

• Design of a system capable of removing multiple noises

• Improved SNR and Perputual Evaluation of Speech Quality(PESQ)

• Simulation of Developed Algorithm

Final Deliverable of the Project Software SystemCore Industry TelecommunicationOther Industries IT Core Technology OthersOther TechnologiesSustainable Development Goals Partnerships to achieve the GoalRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	70000
PC	Equipment	1	70000	70000