Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti...

21
Noise Reduction of Speech Signal using Wavelet Transform with Modified Universal Threshold Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975 – 8887) Volume 20– No.5, April 2011 Presenter Chia-Cheng Chen 1

Transcript of Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti...

Page 1: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Noise Reduction of Speech Signal using Wavelet Transform with Modified Universal

Threshold

Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare

International Journal of Computer Applications (0975 – 8887) Volume 20– No.5, April 2011

Presenter Chia-Cheng Chen 1

Page 2: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Introduction

Multiresolution Analysis Using Filter Bank

Modifier Universal Threshold

Soft and Hard Thresholding

Results and Discussion

Outline

2

Page 3: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Background

3

Page 4: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Discrete-wavelet transform (DWT) based algorithm are used for speech signal denoising. Here both hard and soft thresholding are used for denoising.

Analysis is done on noisy speech signal corrupted

by babble noise at 0dB, 5dB, 10dB and 15dB SNR levels.

Introduction

4

Page 5: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

The DWT uses Multi resolution filter banks and special wavelet filters for the analysis and reconstruction of signals.

It analyse the signal at different frequency bands with different resolutions, decompose the signal into a coarse approximation and detail information.

Multiresolution Analysis Using Filter Bank

5

Page 6: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Two-level wavelet decomposition tree

Multiresolution Analysis Using Filter Bank

6

Page 7: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Two-level wavelet reconstruction tree

Multiresolution Analysis Using Filter Bank

7

Page 8: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Wavelet thresholding is applied to the approximation coefficient (Vn-1) and detail coefficients (Wn,Wn-1).

Multiresolution Analysis Using Filter Bank

8

Page 9: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

In this paper, we removed the babble noise from noisy signal which contain the noise contents of babble noise.

We want to find threshold value that will use to remove noise from noisy signal, but also recover the original signal efficiently.

Modifier Universal Threshold

9

Page 10: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Developed by Donoho and Jonstone and it is called as universal threshold

◦Where N denotes number of samples of noise and is standard deviation of noise.

Modifier Universal Threshold

10

Page 11: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Again Universal threshold was modified with factor “k‟ in order to obtain higher quality output signal:

Modifier Universal Threshold

11

Page 12: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

The soft and hard thresholding methods are used to estimate wavelet coefficients in wavelet threshold denoising.◦Hard thresholding◦ Soft thresholding

Soft and Hard Thresholding

12

Page 13: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Soft and Hard Thresholding

13

Thr=0.4

Z = (-1, 1, 100)

Page 14: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

We implemented babble noise removal algorithm in Matlab 7.10.0 (R2010a).

Speech signal is corrupted by babble noise at 0dB, 5dB, 10dB and 15dB SNR levels.

RESULTS AND DISCUSSION

14

Page 15: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

RESULTS AND DISCUSSION

15

Page 16: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

We choose 5-level DWT and db5 wavelet. Improved threshold value is obtained by replacing threshold “thr‟ (2) with

RESULTS AND DISCUSSION

16

Page 17: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

RESULTS AND DISCUSSION

17

Page 18: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

RESULTS AND DISCUSSION

18

Page 19: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

RESULTS AND DISCUSSION

19

Page 20: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

RESULTS AND DISCUSSION

20

Page 21: Rajeev Aggarwal, Jai Karan Singh, Vijay Kumar Gupta, Sanjay Rathore, Mukesh Tiwari, Dr.Anubhuti Khare International Journal of Computer Applications (0975.

Speech denoising is performed in wavelet domain by thresholding wavelet coefficients.

We found that by using modified universal threshold, we can get the better results of de-noising, especially for low level noise.

RESULTS AND DISCUSSION

21