TUTCRIS - Tampereen teknillinen yliopisto

TUTCRIS

Deep neural network based speech separation optimizing an objective estimator of intelligibility for low latency applications

Tutkimustuotosvertaisarvioitu

Yksityiskohdat

AlkuperäiskieliEnglanti
Otsikko16th International Workshop on Acoustic Signal Enhancement, IWAENC 2018
KustantajaIEEE
Sivut386-390
Sivumäärä5
ISBN (elektroninen)9781538681510
DOI - pysyväislinkit
TilaJulkaistu - 2 marraskuuta 2018
OKM-julkaisutyyppiA4 Artikkeli konferenssijulkaisussa
TapahtumaInternational Workshop on Acoustic Signal Enhancement - Tokyo, Japani
Kesto: 17 syyskuuta 201820 syyskuuta 2018

Conference

ConferenceInternational Workshop on Acoustic Signal Enhancement
MaaJapani
KaupunkiTokyo
Ajanjakso17/09/1820/09/18

Tiivistelmä

Mean square error (MSE) has been the preferred choice as loss function in the current deep neural network (DNN) based speech separation techniques. In this paper, we propose a new cost function with the aim of optimizing the extended short time objective intelligibility (ESTOI) measure. We focus on applications where low algorithmic latency (≤ 10 ms) is important. We use long short-term memory networks (LSTM) and evaluate our proposed approach on four sets of two-speaker mixtures from extended Danish hearing in noise (HINT) dataset. We show that the proposed loss function can offer improved or at par objective intelligibility (in terms of ESTOI) compared to an MSE optimized baseline while resulting in lower objective separation performance (in terms of the source to distortion ratio (SDR)). We then proceed to propose an approach where the network is first initialized with weights optimized for MSE criterion and then trained with the proposed ESTOI loss criterion. This approach mitigates some of the losses in objective separation performance while preserving the gains in objective intelligibility.

Tutkimusalat

Julkaisufoorumi-taso