This work proposes and compares perceptually motivated loss functions for deep learning based binary mask estimation for speech separation. Previous loss functions have focused on maximising classification accuracy of mask estimation but we now propose loss functions that aim to maximise the hit minus false-alarm (HIT-FA) rate which is known to correlate more closely to speech intelligibility. ...

A new real-time two-stage blind source separation (BSS) for convolutive mixtures of speech is proposed, in which a SingleInput Multiple-Output (SIMO)-model-based ICA and binary mask processing are combined. SIMO-model-based ICA can separate the mixed signals, not into monaural source signals but into SIMO-model-based signals from independent sources as they are at the microphones. Thus, the sep...

Multi-speaker separation is necessary to increase intelligibility of speech signals or to improve accuracy of speech recognition systems. Ideal binary mask (IBM) has set a gold standard for speech separation by suppressing the undesired speakers and also by increasing intelligibility of the desired speech. In this work, single frequency filtering (SFF) analysis is used to estimate the mask clos...

In the paper an image interpolation method based on the mathematical morphology is proposed. The method deals with binary images and allows transforming one input binary object into another one. The proposed method is an improvement of the Hausdorff distance approach. Two ways of improvement are proposed. The first one allows controling more precisely the process of object growing during the in...

Deep neural networks (DNNs) are usually used for single channel source separation to predict either soft or binary time frequency masks. The masks are used to separate the sources from the mixed signal. Binary masks produce separated sources with more distortion and less interference than soft masks. In this paper, we propose to use another DNN to combine the estimates of binary and soft masks ...

This study presents a performance comparison of different missing feature imputation techniques under ideal as well as realistic conditions. The particular focus is on recent techniques such as Raj’s soft-decision bounded mean imputation approach and Gemmeke’s sparse imputation. In addition to experiments with oracle masks, we evaluate the usefulness of a number of different mask estimation alg...

Finding optimal data for inpainting is a key problem for imagecompression with partial differential equations. Not only the location of important pixels but also their values should be optimal to maximise the quality gain. The position of important data is usually encoded in a binary mask. Recent studies have shown that allowing non-binary masks may lead to tremendous speedups but comes at the ...

We use the term \Blue Noise Mask" to mean a halftone threshold array that has been constructed to produce pleasing, unstructured (but not white noise) binary patterns. An eecient way to produce a Blue Noise Mask is to construct a set of binary patterns that are constrained by neighboring levels. These binary patterns are appropriate for diierent grey levels, and can be summed to form a single t...

In a previous study, we proposed an alternative masking criterion for binary mask estimation based on the underlying linguistic information. We estimated this mask by selecting from a set of candidate masks at each frame based on the hypotheses from an ASR system. Our previous system provided an 8% reduction in WER. In this work, we present an improved method for selecting the correct candidate...