Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS

Abstract: 

The SpeakerBeam-FE (SBF) method is proposed for speaker extraction. It attempts to overcome the problem of unknown number of speakers in an audio recording during source separation. The mask approximation loss of SBF is sub-optimal, which doesn’t calculate direct signal reconstruction error and consider the speech context. To address these problems, this paper proposes a magnitude and temporal spectrum approximation loss to estimate a phase sensitive mask for the target speaker with the speaker characteristics. Moreover, this paper explores a concatenation framework instead of the context adaptive deep neural network in the SBF method to encode a speaker embedding into the mask estimation network. Experimental results under open evaluation condition show that the proposed method achieves 70.4% and 17.7% relative improvement over the SBF baseline on signal-to-distortion ratio (SDR) and perceptual evaluation of speech quality (PESQ), respectively. A further analysis demonstrates 69.1% and 72.3% relative SDR improvements obtained by the proposed method for different and same gender mixtures.

up
1 user has voted: Chenglin Xu

Paper Details

Authors:
CHENGLIN XU, WEI RAO, ENG SIONG CHNG, HAIZHOU LI
Submitted On:
8 May 2019 - 2:50am
Short Link:
Type:
Poster
Event:
Presenter's Name:
CHENGLIN XU
Paper Code:
SLP-P21.6
Document Year:
2019
Cite

Document Files

ICASSP Poster

(25)

Subscribe

[1] CHENGLIN XU, WEI RAO, ENG SIONG CHNG, HAIZHOU LI, "OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS", IEEE SigPort, 2019. [Online]. Available: http://sigport.org/4028. Accessed: Nov. 18, 2019.
@article{4028-19,
url = {http://sigport.org/4028},
author = {CHENGLIN XU; WEI RAO; ENG SIONG CHNG; HAIZHOU LI },
publisher = {IEEE SigPort},
title = {OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS},
year = {2019} }
TY - EJOUR
T1 - OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS
AU - CHENGLIN XU; WEI RAO; ENG SIONG CHNG; HAIZHOU LI
PY - 2019
PB - IEEE SigPort
UR - http://sigport.org/4028
ER -
CHENGLIN XU, WEI RAO, ENG SIONG CHNG, HAIZHOU LI. (2019). OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS. IEEE SigPort. http://sigport.org/4028
CHENGLIN XU, WEI RAO, ENG SIONG CHNG, HAIZHOU LI, 2019. OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS. Available at: http://sigport.org/4028.
CHENGLIN XU, WEI RAO, ENG SIONG CHNG, HAIZHOU LI. (2019). "OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS." Web.
1. CHENGLIN XU, WEI RAO, ENG SIONG CHNG, HAIZHOU LI. OPTIMIZATION OF SPEAKER EXTRACTION NEURAL NETWORK WITH MAGNITUDE AND TEMPORAL SPECTRUM APPROXIMATION LOSS [Internet]. IEEE SigPort; 2019. Available from : http://sigport.org/4028