Abstract
In this paper, various linear predictive (LP) analysis methods are studied and compared from the points of view ofrobustness to noise and of application to speaker identification. The key to the success of LP techniques is in separatingthe vocal tract information from the pitch information present in a speech signal even under noisy conditions. Inaddition to considering the conventional, one.shot weighted least-squares methods, we propose three other approacheswith the above point as a motivation. The first is an iterative approach that leads to the weighted least absolute valuesolution. The second is an extension of the one-shot least-squares approach and achieves an iterative update of theweights. The update is a function of the residual and is based on minimizing a Mahalanobis distance. Thirdly, theweighted total least-squares formulation is considered. A study of the deviations in the LP parameters was done whennoise (white Gaussian and impulsive) is added to the speech. It was revealed that the most robust method dependson the type of noise. A closed set speaker identification experiment with 20 speakers was conducted using a vectorquantizer classifier trained on clean speech. For a modest codebook size of 32, all of the approaches are comparablewhen the testing condition corresponds to clean speech or speech degraded by white Gaussian noise. When the testinvolves speech degraded by impulse noise, the proposed approach based on minimizing a Mahalanobis distance whichwas found to be the most robust, is also the best for speaker identification.
Original language | English (US) |
---|---|
Pages (from-to) | 83-94 |
Number of pages | 12 |
Journal | Proceedings of SPIE - The International Society for Optical Engineering |
Volume | 2277 |
DOIs | |
State | Published - Oct 25 1994 |
Externally published | Yes |
Event | Automatic Systems for the Identification and Inspection of Humans 1994 - San Diego, United States Duration: Jul 24 1994 → Jul 29 1994 |
All Science Journal Classification (ASJC) codes
- Electronic, Optical and Magnetic Materials
- Condensed Matter Physics
- Computer Science Applications
- Applied Mathematics
- Electrical and Electronic Engineering