TY - JOUR
T1 - Correlation distance skip connection denoising autoencoder (CDSK-DAE) for speech feature enhancement
AU - Badi, Alzahra
AU - Park, Sangwook
AU - Han, David K.
AU - Ko, Hanseok
N1 - Funding Information:
The work was supported by the Korea Environmental Industry and Technology Institute’s environmental policy-based public technology development project ( 2017000210001 ) that is funded by the Ministry of Environment. The contribution of David K. Han was supported by the U.S. Army Research Laboratory.
PY - 2020/6
Y1 - 2020/6
N2 - Performance of learning based Automatic Speech Recognition (ASR) is susceptible to noise, especially when it is introduced in the testing data while not presented in the training data. This work focuses on a feature enhancement for noise robust end-to-end ASR system by introducing a novel variant of denoising autoencoder (DAE). The proposed method uses skip connections in both encoder and decoder sides by passing speech information of the target frame from input to the model. It also uses a new objective function in training model that uses a correlation distance measure in penalty terms by measuring dependency of the latent target features and the model (latent features and enhanced features obtained from the DAE). Performance of the proposed method was compared against a conventional model and a state of the art model under both seen and unseen noisy environments of 7 different types of background noise with different SNR levels (0, 5, 10 and 20 dB). The proposed method also is tested using linear and non-linear penalty terms as well, where, they both show an improvement on the overall average WER under noisy conditions both seen and unseen in comparison to the state-of-the-art model.
AB - Performance of learning based Automatic Speech Recognition (ASR) is susceptible to noise, especially when it is introduced in the testing data while not presented in the training data. This work focuses on a feature enhancement for noise robust end-to-end ASR system by introducing a novel variant of denoising autoencoder (DAE). The proposed method uses skip connections in both encoder and decoder sides by passing speech information of the target frame from input to the model. It also uses a new objective function in training model that uses a correlation distance measure in penalty terms by measuring dependency of the latent target features and the model (latent features and enhanced features obtained from the DAE). Performance of the proposed method was compared against a conventional model and a state of the art model under both seen and unseen noisy environments of 7 different types of background noise with different SNR levels (0, 5, 10 and 20 dB). The proposed method also is tested using linear and non-linear penalty terms as well, where, they both show an improvement on the overall average WER under noisy conditions both seen and unseen in comparison to the state-of-the-art model.
KW - Automatic speech recognition (ASR)
KW - Correlation distance measure (CDM)
KW - Skip connection Denoising Autoencoder (SK-DAE)
UR - http://www.scopus.com/inward/record.url?scp=85078119404&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85078119404&partnerID=8YFLogxK
U2 - 10.1016/j.apacoust.2020.107213
DO - 10.1016/j.apacoust.2020.107213
M3 - Article
AN - SCOPUS:85078119404
VL - 163
JO - Applied Acoustics
JF - Applied Acoustics
SN - 0003-682X
M1 - 107213
ER -