Spectral Features Based on Local Normalized Center Moments for Speech Emotion Recognition

Huawei TAO; Ruiyu LIANG; Xinran ZHANG; Li ZHAO

doi:10.1587/transfun.E99.A.1863

Spectral Features Based on Local Normalized Center Moments for Speech Emotion Recognition

Huawei TAO, Ruiyu LIANG, Xinran ZHANG, Li ZHAO

Full Text Views

0

Cite this

Summary :

To discuss whether rotational invariance is the main role in spectrogram features, new spectral features based on local normalized center moments, denoted by LNCMSF, are proposed. The proposed LNCMSF firstly adopts 2nd order normalized center moments to describe local energy distribution of the logarithmic energy spectrum, then normalized center moment spectrograms NC₁ and NC₂ are gained. Secondly, DCT (Discrete Cosine Transform) is used to eliminate the correlation of NC₁ and NC₂, then high order cepstral coefficients TNC₁ and TNC₂ are obtained. Finally, LNCMSF is generated by combining NC₁, NC₂, TNC₁ and TNC₂. The rotational invariance test experiment shows that the rotational invariance is not a necessary property in partial spectrogram features. The recognition experiment shows that the maximum UA (Unweighted Average of Class-Wise Recall Rate) of LNCMSF are improved by at least 10.7% and 1.2% respectively, compared to that of MFCC (Mel Frequency Cepstrum Coefficient) and HuWSF (Weighted Spectral Features Based on Local Hu Moments).

Publication: IEICE TRANSACTIONS on Fundamentals Vol.E99-A No.10 pp.1863-1866

Publication Date: 2016/10/01

Publicized

Online ISSN: 1745-1337

DOI: 10.1587/transfun.E99.A.1863

Type of Manuscript: LETTER

Category: Speech and Hearing

Authors

Huawei TAO
  Southeast University
Ruiyu LIANG
  Southeast University,Nanjing Institute of Technology
Xinran ZHANG
  Southeast University
Li ZHAO
  Southeast University

Keyword

normalized center moments, Hu invariant moments, speech emotion recognition, spectral features

Cite this

Copy

Huawei TAO, Ruiyu LIANG, Xinran ZHANG, Li ZHAO, "Spectral Features Based on Local Normalized Center Moments for Speech Emotion Recognition" in IEICE TRANSACTIONS on Fundamentals, vol. E99-A, no. 10, pp. 1863-1866, October 2016, doi: 10.1587/transfun.E99.A.1863.
Abstract: To discuss whether rotational invariance is the main role in spectrogram features, new spectral features based on local normalized center moments, denoted by LNCMSF, are proposed. The proposed LNCMSF firstly adopts 2nd order normalized center moments to describe local energy distribution of the logarithmic energy spectrum, then normalized center moment spectrograms NC₁ and NC₂ are gained. Secondly, DCT (Discrete Cosine Transform) is used to eliminate the correlation of NC₁ and NC₂, then high order cepstral coefficients TNC₁ and TNC₂ are obtained. Finally, LNCMSF is generated by combining NC₁, NC₂, TNC₁ and TNC₂. The rotational invariance test experiment shows that the rotational invariance is not a necessary property in partial spectrogram features. The recognition experiment shows that the maximum UA (Unweighted Average of Class-Wise Recall Rate) of LNCMSF are improved by at least 10.7% and 1.2% respectively, compared to that of MFCC (Mel Frequency Cepstrum Coefficient) and HuWSF (Weighted Spectral Features Based on Local Hu Moments).
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.E99.A.1863/_p

Copy

@ARTICLE{e99-a_10_1863,
author={Huawei TAO, Ruiyu LIANG, Xinran ZHANG, Li ZHAO, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Spectral Features Based on Local Normalized Center Moments for Speech Emotion Recognition},
year={2016},
volume={E99-A},
number={10},
pages={1863-1866},
abstract={To discuss whether rotational invariance is the main role in spectrogram features, new spectral features based on local normalized center moments, denoted by LNCMSF, are proposed. The proposed LNCMSF firstly adopts 2nd order normalized center moments to describe local energy distribution of the logarithmic energy spectrum, then normalized center moment spectrograms NC₁ and NC₂ are gained. Secondly, DCT (Discrete Cosine Transform) is used to eliminate the correlation of NC₁ and NC₂, then high order cepstral coefficients TNC₁ and TNC₂ are obtained. Finally, LNCMSF is generated by combining NC₁, NC₂, TNC₁ and TNC₂. The rotational invariance test experiment shows that the rotational invariance is not a necessary property in partial spectrogram features. The recognition experiment shows that the maximum UA (Unweighted Average of Class-Wise Recall Rate) of LNCMSF are improved by at least 10.7% and 1.2% respectively, compared to that of MFCC (Mel Frequency Cepstrum Coefficient) and HuWSF (Weighted Spectral Features Based on Local Hu Moments).},
keywords={},
doi={10.1587/transfun.E99.A.1863},
ISSN={1745-1337},
month={October},}

Copy

TY - JOUR
TI - Spectral Features Based on Local Normalized Center Moments for Speech Emotion Recognition
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 1863
EP - 1866
AU - Huawei TAO
AU - Ruiyu LIANG
AU - Xinran ZHANG
AU - Li ZHAO
PY - 2016
DO - 10.1587/transfun.E99.A.1863
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E99-A
IS - 10
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - October 2016
AB - To discuss whether rotational invariance is the main role in spectrogram features, new spectral features based on local normalized center moments, denoted by LNCMSF, are proposed. The proposed LNCMSF firstly adopts 2nd order normalized center moments to describe local energy distribution of the logarithmic energy spectrum, then normalized center moment spectrograms NC₁ and NC₂ are gained. Secondly, DCT (Discrete Cosine Transform) is used to eliminate the correlation of NC₁ and NC₂, then high order cepstral coefficients TNC₁ and TNC₂ are obtained. Finally, LNCMSF is generated by combining NC₁, NC₂, TNC₁ and TNC₂. The rotational invariance test experiment shows that the rotational invariance is not a necessary property in partial spectrogram features. The recognition experiment shows that the maximum UA (Unweighted Average of Class-Wise Recall Rate) of LNCMSF are improved by at least 10.7% and 1.2% respectively, compared to that of MFCC (Mel Frequency Cepstrum Coefficient) and HuWSF (Weighted Spectral Features Based on Local Hu Moments).
ER -