Speech Enhancement with Impact Noise Activity Detection Based on the Kurtosis of an Instantaneous Power Spectrum

Naoto SASAOKA; Naoya HAMAHASHI; Yoshio ITOH

doi:10.1587/transfun.E100.A.1942

Speech Enhancement with Impact Noise Activity Detection Based on the Kurtosis of an Instantaneous Power Spectrum

Naoto SASAOKA, Naoya HAMAHASHI, Yoshio ITOH

Full Text Views

0

Share
Cite this

Summary :

In a speech enhancement system for impact noise, it is important for any impact noise activity to be detected. However, because impact noise occurs suddenly, it is not always easy to detect. We propose a method for impact noise activity detection based on the kurtosis of an instantaneous power spectrum. The continuous duration of a generalized impact noise is shorter than that of speech, and the power of such impact noise varies dramatically. Consequently, the distribution of the instantaneous power spectrum of impact noise is different from that of speech. The proposed detection takes advantage of kurtosis, which depends on the sharpness and skirt of the distribution. Simulation results show that the proposed noise activity detection improves the performance of the speech enhancement system.

Publication: IEICE TRANSACTIONS on Fundamentals Vol.E100-A No.9 pp.1942-1950

Publication Date: 2017/09/01

Publicized

Online ISSN: 1745-1337

DOI: 10.1587/transfun.E100.A.1942

Type of Manuscript: PAPER

Category: Digital Signal Processing

Authors

Naoto SASAOKA
  Tottori University
Naoya HAMAHASHI
  Tottori University
Yoshio ITOH
  Tottori University

Keyword

speech enhancement, noise activity detection, impact noise, kurtosis, high-order cumulant

Cite this

Copy

Naoto SASAOKA, Naoya HAMAHASHI, Yoshio ITOH, "Speech Enhancement with Impact Noise Activity Detection Based on the Kurtosis of an Instantaneous Power Spectrum" in IEICE TRANSACTIONS on Fundamentals, vol. E100-A, no. 9, pp. 1942-1950, September 2017, doi: 10.1587/transfun.E100.A.1942.
Abstract: In a speech enhancement system for impact noise, it is important for any impact noise activity to be detected. However, because impact noise occurs suddenly, it is not always easy to detect. We propose a method for impact noise activity detection based on the kurtosis of an instantaneous power spectrum. The continuous duration of a generalized impact noise is shorter than that of speech, and the power of such impact noise varies dramatically. Consequently, the distribution of the instantaneous power spectrum of impact noise is different from that of speech. The proposed detection takes advantage of kurtosis, which depends on the sharpness and skirt of the distribution. Simulation results show that the proposed noise activity detection improves the performance of the speech enhancement system.
URL: https://globals.ieice.org/en_transactions/fundamentals/10.1587/transfun.E100.A.1942/_p

Copy

@ARTICLE{e100-a_9_1942,
author={Naoto SASAOKA, Naoya HAMAHASHI, Yoshio ITOH, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Speech Enhancement with Impact Noise Activity Detection Based on the Kurtosis of an Instantaneous Power Spectrum},
year={2017},
volume={E100-A},
number={9},
pages={1942-1950},
abstract={In a speech enhancement system for impact noise, it is important for any impact noise activity to be detected. However, because impact noise occurs suddenly, it is not always easy to detect. We propose a method for impact noise activity detection based on the kurtosis of an instantaneous power spectrum. The continuous duration of a generalized impact noise is shorter than that of speech, and the power of such impact noise varies dramatically. Consequently, the distribution of the instantaneous power spectrum of impact noise is different from that of speech. The proposed detection takes advantage of kurtosis, which depends on the sharpness and skirt of the distribution. Simulation results show that the proposed noise activity detection improves the performance of the speech enhancement system.},
keywords={},
doi={10.1587/transfun.E100.A.1942},
ISSN={1745-1337},
month={September},}

Copy

TY - JOUR
TI - Speech Enhancement with Impact Noise Activity Detection Based on the Kurtosis of an Instantaneous Power Spectrum
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 1942
EP - 1950
AU - Naoto SASAOKA
AU - Naoya HAMAHASHI
AU - Yoshio ITOH
PY - 2017
DO - 10.1587/transfun.E100.A.1942
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E100-A
IS - 9
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - September 2017
AB - In a speech enhancement system for impact noise, it is important for any impact noise activity to be detected. However, because impact noise occurs suddenly, it is not always easy to detect. We propose a method for impact noise activity detection based on the kurtosis of an instantaneous power spectrum. The continuous duration of a generalized impact noise is shorter than that of speech, and the power of such impact noise varies dramatically. Consequently, the distribution of the instantaneous power spectrum of impact noise is different from that of speech. The proposed detection takes advantage of kurtosis, which depends on the sharpness and skirt of the distribution. Simulation results show that the proposed noise activity detection improves the performance of the speech enhancement system.
ER -