Singh, S. and Assaf, Mansour (2016) A perfect balance of sparsity and acoustic hole in speech signal and its application in speaker recognition system. Middle-East Journal of Scientific Research, 24 (11). pp. 3527-3541. ISSN 1990-9233
Full text not available from this repository. (Request a copy)Abstract
This paper addresses the issue of balancing the acoustic hole and sparsing the speech signal
enrollment for training and testing in Automatic Speaker Recognition (ASR) system. Sparsing techniques involve the representation of a small number of coefficients that hold a large amount of the energy. Sparsity can play a major role in resolving the issue of dealing with big data in ASR by applying speech compression techniques and information storage in databases. Spectral domain compression of the speech signal using novel sparsing algorithms that balance the sparsity of speech signal with the acoustic hole is proposed. The speech signal is converted to a spectral domain using the Discrete Rajan Transform (DRT) and only first and mid-spectrum component in each block of size 8x1 retained forcing the remaining component to zero. The speech signal spectrum can be maximally compressed at 8:1 ratio to the unique one with balancing acoustic hole and synthesized speech signal, which can be used in ASR systems. A balanced spectrally compressed speech signal can be stored in database as a speaker representative and during training and testing
time it can be synthesized using the Inverse Discrete Rajan Transform (IDRT). Simulation results, shows acceptable speech signal spectral compression that balances sparsity and the generation of the acoustic hole
is 75% with 94.8% efficiency without sparsification and 99.1% efficiency with TIMIT database respectively.
| Item Type: | Journal Article |
|---|---|
| Subjects: | T Technology > T Technology (General) |
| Divisions: | Faculty of Science, Technology and Environment (FSTE) > School of Engineering and Physics |
| Depositing User: | Mansour Assaf |
| Date Deposited: | 03 Jan 2017 02:26 |
| Last Modified: | 03 Jan 2017 02:26 |
| URI: | https://repository.usp.ac.fj/id/eprint/9550 |
Actions (login required)
 |
View Item |