A Comparison of Generalized Cross-Correlation Methods for Time Delay Estimation
Article Details
Pub. Date
:
Nov,
2016
Product Name
:
The IUP Journal of Telecommunications
Product Type
:
Article
Product Code
:
IJTC31611
Author Name
:
Ritu and Sanjeev Kumar Dhull
Availability
:
YES
Subject/Domain
:
Science & Technology
Download Format
:
PDF Format
No. of Pages
:
16
Price
For delivery in electronic
format: Rs. 50;
For delivery through courier (within India): Rs.
50 + Rs. 25 for Shipping & Handling Charges
Download
To download this Article click on the button below:
Abstract
Time Delay Estimation (TDE) between the signals received at the spatial sensor pair has been a topic of significant importance because this estimated delay can be used as input in plenty of applications for further processing of the signal. The estimation of the delay should be accurate enough in order to avoid or minimize the propagation of the error to the subsequent stages. The generalized cross-correlation method is the most popular and widely used technique for TDE due to accuracy and low computational complexity. The paper presents a comparison of weighting functions used in generalized cross-correlation along with their simulation results. It has been shown how each weight affects the peak sharpening in the correlation function. Simulation has been carried out with different types of sound and noise. The effect of varying Signal-to-Noise Ratio (SNR) on the estimated delay has also been studied.
Description
Assume a sound signal originates from a source at certain time; it arrives at different microphones at different intervals depending upon their locations. This estimated time delay is processed further to extract the desired information and therefore the accuracy of the signal being processed depends on this preliminary step. It is widely used in applications such as passive source detection, identification, localization, speech enhancement, speech recognition and many more. This task of estimating delay is very challenging because of the non-stationary nature of speech, coherent noise and of room acoustic reverberation (Patel and Shiva, 2015). Peak in the correlation function gets weakened with low Signal-to-Noise Ratio (SNR) and strong reverberation. Not only this, there occurs multiple peaks in correlation function due to strong echoes and hence the desired estimation could not be extracted.
Keywords
Telecommunications Journal, Phase Transform (PHAT), Smoothed Coherence Transform (SCOT), ROTH, Maximum Likelihood (ML), Hassan and Boucher (HB), Cross Correlation (CC), Generalized Cross Correlation (GCC), Time Delay Estimation (TDE).