/acoustic-interference-cancellation

acoustic interference (echo) cancellation project in summer internship

Primary LanguageMATLAB

Acoustic Interference Cancellation

Introduction

This is the GitHub page for Acoustic Interference (Echo) Cancellation, which is the project in the summer internship in Shenzhen Micro and Nano Institute.

The main contributor to this project is:
Zhuowen Lin
The programming language is MATLAB.

Mind map

mind map

Repository document components

  • Daily Logs: Records the daily schedules.
  • Fullband processing: Codes for fullband adaptive acoustic interference cancellation.
  • Speex Codec: MATLAB version of the acoustic echo cancellation program in Speex Codec. The code is forked from here.
  • Subband processing: Codes for subband adaptive acoustic interference cancellation.
    • Book: MATLAB codes for different algorithms proposed in different chapters of the book Subband Adaptive Filtering. (The pdf version of this book is included in the document Supporting Materials.)
    • Common Code: All the usable codes in the book Subband Adaptive Filtering.
  • Supporting materials: pdf version of supporting materials including books and papers.

Result demonstration

The spectrogram of the original acoustic file, which contains background musical interference signal and demanded speech signal "Xiao Du Xiao Du" under 500 Hz.
Original spectrogram
The spectrogram of the subband filtering processed acoustic file, which contains deteriorated and suppressed background musical interference signal and clearer demanded speech signal "Xiao Du Xiao Du" under 500 Hz. Subband filtering processed spectrogram
The result can be even better by first utilizing the DNN-Speech-enhancement-demo-tool and then doing subband filtering. The basic idea is to use DNN-Speech-enhancement-demo-tool to clear the background musical interference effectively and use subband filtering to clear the human singing voice interference. The resulting spectrogram is shown below. DNN+Subband filtering processed spectrogram