DNN-based enhancement of noisy and reverberant speech

Zhao, Yan; Wang, DeLiang; Merks, Ivo; Zhang, Tao

doi:10.1109/icassp.2016.7472934

Cited by 69 publications

(33 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, we use masking based source separation paradigm (e.g., used in [4,25,26]) where a DNN is used to predict a time-frequency mask corresponding to a target speaker. The ESTOI computation however is done for estimated and reference source spectra.…”

Section: Implicit Time-frequency Maskingmentioning

confidence: 99%

Deep Neural Network Based Speech Separation Optimizing an Objective Estimator of Intelligibility for Low Latency Applications

Naithani

Nikunen

Bramsløw

et al. 2018

2018 16th International Workshop on Acoustic Signal Enhancement (IWAENC)

View full text Add to dashboard Cite

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.

show abstract

Section: Implicit Time-frequency Maskingmentioning

confidence: 99%

Deep Neural Network Based Speech Separation Optimizing an Objective Estimator of Intelligibility for Low Latency Applications

Naithani

Nikunen

Bramsløw

et al. 2018

2018 16th International Workshop on Acoustic Signal Enhancement (IWAENC)

View full text Add to dashboard Cite

show abstract

“…The activation functions for the T-F mask φ g , variance φ σ , and hidden units φ h were the sigmoid function, exponential function, and rectified linear unit (ReLU), respectively. The context window size was Q = 5, and the variance regularization parameter in (15) was C σ = 10 −41 . The Adam method [47] was used as a gradient method.…”

Section: Methodsmentioning

confidence: 99%

DNN-Based Source Enhancement to Increase Objective Sound Quality Assessment Score

Koizumi

Niwa

Hioka

et al. 2018

IEEE/ACM Trans. Audio Speech Lang. Process.

View full text Add to dashboard Cite

We propose a training method for deep neural network (DNN)-based source enhancement to increase objective sound quality assessment (OSQA) scores such as the perceptual evaluation of speech quality (PESQ). In many conventional studies, DNNs have been used as a mapping function to estimate time-frequency masks and trained to minimize an analytically tractable objective function such as the mean squared error (MSE). Since OSQA scores have been used widely for soundquality evaluation, constructing DNNs to increase OSQA scores would be better than using the minimum-MSE to create highquality output signals. However, since most OSQA scores are not analytically tractable, i.e., they are black boxes, the gradient of the objective function cannot be calculated by simply applying back-propagation. To calculate the gradient of the OSQA-based objective function, we formulated a DNN optimization scheme on the basis of black-box optimization, which is used for training a computer that plays a game. For a black-box-optimization scheme, we adopt the policy gradient method for calculating the gradient on the basis of a sampling algorithm. To simulate output signals using the sampling algorithm, DNNs are used to estimate the probability-density function of the output signals that maximize OSQA scores. The OSQA scores are calculated from the simulated output signals, and the DNNs are trained to increase the probability of generating the simulated output signals that achieve high OSQA scores. Through several experiments, we found that OSQA scores significantly increased by applying the proposed method, even though the MSE was not minimized.

show abstract

“…Secondly, a sampling method needs to be defined for eqns. (9) and (11), that incorporates both clean and noisy inputs. To do this, a pair of corresponding clean and noisy spectra sequences, z, are sampled, from which G generates a fake samplex t .…”

Section: Discriminator Architecturementioning

confidence: 99%