International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395 -0056
Volume: 04 Issue: 03 | Mar -2017
p-ISSN: 2395-0072
www.irjet.net
Speech Enhancement using Compressive Sensing K.Kiruthiga1, J.Indra2 1PG 2Assistant
Scholar,Dept. Of EIE, Kongu Engineering College, Tamilnadu,India Professor (SLG),Dept. Of EIE, Kongu Engineering College, Tamilnadu, India
---------------------------------------------------------------------***-----------------------------------------------------------------
Abstract - Speech enhancement is a technique which is used to reduce the background noise present in the speech signal. The noises are additive noise, echo, reverberation and speaker interference. The aim of the proposed method is to reduce the background noise present in the speech signal by using compressive sensing. The goal of compressive sensing is to compress the speech signal at transmitter and decompress it at the receiver from far less samples than the nyquist rate. In this work, a speech signal is taken and then it is compressively sampled using a measurement matrix which in case is composed of randomly generated numbers. The output of the compressed sensing algorithm is the observation vector which is transmitted to the receiver. At the receiver section, signal is reconstructed from a significant small numbers of samples by using l1- minimization. MATLAB simulations are performed to compress the speech signal below the nyquist rate and to reconstruct it without losing any important information.
2. Compressive sensing
Compressive sensing involves recovering the speech signal from far less samples than the nyquist rate [8]. Fig.1 shows the basic block diagram of compressive sensing. Initially, the signal is sampled using nyquist rate, whereas with the help of compressive sensing the signal is sampled below the nyquist rate [5]. The signal is transformed into a domain in which it shows sparse representation. Then the signal is transmitted and stored in the channel by the receiver side [13].
Key Words: Speech enhancement, Compressive sensing, DCT, l1 –minimization, Measurement matrix.
1. INTRODUCTION In recent years, various signal sampling schemes have been developed. However, such sampling methods are difficult to implement. So before sampling the signal it should have sufficient information about the reconstruction kernel. The emerging compressive sensing theory shows that an unevenly sampled discrete signal can be perfectly reconstructed by high probability of success by using different optimization techniques and by considering fewer random projections or measurements compared to the Nyquist standard. Amart Sulong et al proposed the compressive sensing method by combining randomized measurement matrix with the wiener filter to reduce the noisy speech signal and thereby producing high signal to noise ratio [1]. Joel A. Tropp et al demonstrated the theoretical and empirical work of Orthogonal Matching Pursuit (OMP) which is effective alternative to (BP) for signal recovery from random measurements [2]. Phu Ngoc Le et al proposed an improved soft – thresholding method for DCT speech enhancement [3]. Vahid Abolghasemi focused on proper estimation of measurement matrix for compressive sampling of the signal [4].
Š 2017, IRJET
|
Impact Factor value: 5.181
|
Fig.1 Basic block diagram of compressive sensing Finally the signal is reconstructed from the samples by using one of the different optimization techniques available.
3. Noizeus Corpus Thirty sentences from the IEEE sentence database (IEEE Subcommittee 1969) were recorded in a sound-proof booth using Tucker Davis Technologies (TDT) recording equipment [12]. The sentences were produced by three male and three female speakers. The sentences were originally sampled at 25 kHz and down sampled to 8 kHz and eight basic noise signals under different environmental conditions are taken from the AURORA database [9]. It has the recordings from different places like Babble, Car, Exhibition hall, Restaurant, Street, Airport, Train station and Train.
ISO 9001:2008 Certified Journal
|
Page 1125