### 抄録

This letter presents a new algorithm for blind dereverberation and echo cancellation based on independent component analysis (ICA) for actual acoustic signals. We focus on frequency domain ICA (FD-ICA) because its computational cost and speed of learning convergence are sufficiently reasonable for practical applications such as hands-free speech recognition. In applying conventional FD-ICA as a preprocessing of automatic speech recognition in noisy environments, one of the most critical problems is how to copewith reverberations. To extract a clean signal from the reverberant observation, we model the separation process in the shorttime Fourier transform domain and apply the multiple input/output inverse-filtering theorem (MINT) to the FD-ICA separation model. A naive implementation of this method is computationally expensive, because its time complexity is the second order of reverberation time. Therefore, themain issue in dereverberation is to reduce the high computational cost of ICA. In this letter, wereduce the computational complexity to the linear order of the reverberation time by using two techniques: (1) a separation model based on the independence of delayed observed signals with MINT and (2) spatial sphering for preprocessing. Experiments show that the computational cost grows in proportion to the linear order of the reverberation time and that ourmethod improves the word correctness of automatic speech recognition by 10 to 20 points in a RT_{20} =670 ms reverberant environment.

元の言語 | English |
---|---|

ページ（範囲） | 234-272 |

ページ数 | 39 |

ジャーナル | Neural Computation |

巻 | 24 |

発行部数 | 1 |

DOI | |

出版物ステータス | Published - 2012 |

外部発表 | Yes |

### Fingerprint

### ASJC Scopus subject areas

- Cognitive Neuroscience
- Arts and Humanities (miscellaneous)

### これを引用

*Neural Computation*,

*24*(1), 234-272. https://doi.org/10.1162/NECO_a_00219

**Efficient blind dereverberation and echo cancellation based on independent component analysis for actual acoustic signals.** / Takeda, Ryu; Nakadai, Kazuhiro; Takahashi, Toru; Komatani, Kazunori; Ogata, Tetsuya; Okuno, Hiroshi G.

研究成果: Article

*Neural Computation*, 巻. 24, 番号 1, pp. 234-272. https://doi.org/10.1162/NECO_a_00219

}

TY - JOUR

T1 - Efficient blind dereverberation and echo cancellation based on independent component analysis for actual acoustic signals

AU - Takeda, Ryu

AU - Nakadai, Kazuhiro

AU - Takahashi, Toru

AU - Komatani, Kazunori

AU - Ogata, Tetsuya

AU - Okuno, Hiroshi G.

PY - 2012

Y1 - 2012

N2 - This letter presents a new algorithm for blind dereverberation and echo cancellation based on independent component analysis (ICA) for actual acoustic signals. We focus on frequency domain ICA (FD-ICA) because its computational cost and speed of learning convergence are sufficiently reasonable for practical applications such as hands-free speech recognition. In applying conventional FD-ICA as a preprocessing of automatic speech recognition in noisy environments, one of the most critical problems is how to copewith reverberations. To extract a clean signal from the reverberant observation, we model the separation process in the shorttime Fourier transform domain and apply the multiple input/output inverse-filtering theorem (MINT) to the FD-ICA separation model. A naive implementation of this method is computationally expensive, because its time complexity is the second order of reverberation time. Therefore, themain issue in dereverberation is to reduce the high computational cost of ICA. In this letter, wereduce the computational complexity to the linear order of the reverberation time by using two techniques: (1) a separation model based on the independence of delayed observed signals with MINT and (2) spatial sphering for preprocessing. Experiments show that the computational cost grows in proportion to the linear order of the reverberation time and that ourmethod improves the word correctness of automatic speech recognition by 10 to 20 points in a RT20 =670 ms reverberant environment.

AB - This letter presents a new algorithm for blind dereverberation and echo cancellation based on independent component analysis (ICA) for actual acoustic signals. We focus on frequency domain ICA (FD-ICA) because its computational cost and speed of learning convergence are sufficiently reasonable for practical applications such as hands-free speech recognition. In applying conventional FD-ICA as a preprocessing of automatic speech recognition in noisy environments, one of the most critical problems is how to copewith reverberations. To extract a clean signal from the reverberant observation, we model the separation process in the shorttime Fourier transform domain and apply the multiple input/output inverse-filtering theorem (MINT) to the FD-ICA separation model. A naive implementation of this method is computationally expensive, because its time complexity is the second order of reverberation time. Therefore, themain issue in dereverberation is to reduce the high computational cost of ICA. In this letter, wereduce the computational complexity to the linear order of the reverberation time by using two techniques: (1) a separation model based on the independence of delayed observed signals with MINT and (2) spatial sphering for preprocessing. Experiments show that the computational cost grows in proportion to the linear order of the reverberation time and that ourmethod improves the word correctness of automatic speech recognition by 10 to 20 points in a RT20 =670 ms reverberant environment.

UR - http://www.scopus.com/inward/record.url?scp=83455200229&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=83455200229&partnerID=8YFLogxK

U2 - 10.1162/NECO_a_00219

DO - 10.1162/NECO_a_00219

M3 - Article

C2 - 22023192

AN - SCOPUS:83455200229

VL - 24

SP - 234

EP - 272

JO - Neural Computation

JF - Neural Computation

SN - 0899-7667

IS - 1

ER -