Fourier Domain Speech Enhancement by Constrained Optimization and Temporal Masking

Author

Jin, Wen

Author_Institution

Dept. of Electr. & Comput. Eng., Univ. of Miami, Coral Gables, FL

fYear

2006

fDate

Sept. 2006

Firstpage

585

Lastpage

589

Abstract

A speech enhancement algorithm is proposed that exploits the masking properties of the human auditory system. A clean speech estimator is derived in such a way that both the signal distortion and residual noise are kept inaudible. Temporal as well as simultaneous masking effects are incorporated in estimation of the masking thresholds. The proposed speech enhancement algorithm is evaluated by log spectral distortion and ITU PESQ scores. Experimental results indicate that the proposed method out-performs a former psychoacoustically motivated speech enhancement algorithm

Keywords

Fourier analysis; distortion; noise; speech enhancement; Fourier domain speech enhancement; human auditory system; log spectral distortion; noise; signal distortion; temporal masking; Additive noise; Auditory system; Constraint optimization; Discrete cosine transforms; Fourier transforms; Humans; Masking threshold; Psychoacoustic models; Psychology; Speech enhancement; Speech enhancement; psychoacoustical model; temporal masking;

fLanguage

English

Publisher

ieee

Conference_Titel

Digital Signal Processing Workshop, 12th - Signal Processing Education Workshop, 4th

Conference_Location

Teton National Park, WY

Print_ISBN

1-4244-3534-3

Electronic_ISBN

1-4244-0535-1

Type

conf

DOI

10.1109/DSPWS.2006.265491

Filename

4041132