| Presentation | 2014-01-23 Speaker recognition based on log-linear models using feature generation by variational Bayesian method Akifumi TSUGE, Kei HASHIMOTO, Yoshihiko NANKAKU, Keiichi TOKUDA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | This paper presents a speaker recognition technique based on log-linear models (LLMs) using Bayesian statistics. Since discriminative models can use various features in the unified framework, preparation of features that are useful for classification is an important problem. Statistics obtained from Gaussian Mixture Models (GMMs) trained by the maximum likelihood method or the maximum a posteriori method are recently used as features for speaker recognition. However, these training methods often occur the over-fitting problem. In this paper, the Bayesian approach is applied to train GMMs and statistics of GMMs in the Bayesian approach are used as features of LLMs. Experimental results show that the proposed LLM-based method significantly improved the identification rates from conventional GMM-based methods. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | speaker recognition / GMM / Bayesian approach / log-linear model |
| Paper # | SP2013-98 |
| Date of Issue |
| Conference Information | |
| Committee | SP |
|---|---|
| Conference Date | 2014/1/16(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Speech (SP) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Speaker recognition based on log-linear models using feature generation by variational Bayesian method |
| Sub Title (in English) | |
| Keyword(1) | speaker recognition |
| Keyword(2) | GMM |
| Keyword(3) | Bayesian approach |
| Keyword(4) | log-linear model |
| 1st Author's Name | Akifumi TSUGE |
| 1st Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology() |
| 2nd Author's Name | Kei HASHIMOTO |
| 2nd Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology |
| 3rd Author's Name | Yoshihiko NANKAKU |
| 3rd Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology |
| 4th Author's Name | Keiichi TOKUDA |
| 4th Author's Affiliation | Department of Computer Science and Engineering, Nagoya Institute of Technology |
| Date | 2014-01-23 |
| Paper # | SP2013-98 |
| Volume (vol) | vol.113 |
| Number (no) | 404 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |