Kateřina Žmolíková

Ján Švec, Kateřina Žmolíková, Martin Kocour et al.

Recently, the performance of blind speech separation (BSS) and target speech extraction (TSE) has greatly progressed. Most works, however, focus on relatively well-controlled conditions using, e.g., read speech. The performance may degrade in more realistic situations. One of the factors causing such degradation may be intrinsic speaker variability, such as emotions, occurring commonly in realistic speech. In this paper, we investigate the influence of emotions on TSE and BSS. We create a new test dataset of emotional mixtures for the evaluation of TSE and BSS. This dataset combines LibriSpeech and Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS). Through controlled experiments, we can analyze the impact of different emotions on the performance of BSS and TSE. We observe that BSS is relatively robust to emotions, while TSE, which requires identifying and extracting the speech of a target speaker, is much more sensitive to emotions. On comparative speaker verification experiments we show that identifying the target speaker may be particularly challenging when dealing with emotional speech. Using our findings, we outline potential future directions that could improve the robustness of BSS and TSE systems toward emotional speech.

Martin Kocour, Kateřina Žmolíková, Lucas Ondel et al.

In typical multi-talker speech recognition systems, a neural network-based acoustic model predicts senone state posteriors for each speaker. These are later used by a single-talker decoder which is applied on each speaker-specific output stream separately. In this work, we argue that such a scheme is sub-optimal and propose a principled solution that decodes all speakers jointly. We modify the acoustic model to predict joint state posteriors for all speakers, enabling the network to express uncertainty about the attribution of parts of the speech signal to the speakers. We employ a joint decoder that can make use of this uncertainty together with higher-level language information. For this, we revisit decoding algorithms used in factorial generative models in early multi-talker speech recognition systems. In contrast with these early works, we replace the GMM acoustic model with DNN, which provides greater modeling power and simplifies part of the inference. We demonstrate the advantage of joint decoding in proof of concept experiments on a mixed-TIDIGITS dataset.

Marc Delcroix, Katerina Zmolikova, Tsubasa Ochiai et al.

Target speech extraction, which extracts the speech of a target speaker in a mixture given auxiliary speaker clues, has recently received increased interest. Various clues have been investigated such as pre-recorded enrollment utterances, direction information, or video of the target speaker. In this paper, we explore the use of speaker activity information as an auxiliary clue for single-channel neural network-based speech extraction. We propose a speaker activity driven speech extraction neural network (ADEnet) and show that it can achieve performance levels competitive with enrollment-based approaches, without the need for pre-recordings. We further demonstrate the potential of the proposed approach for processing meeting-like recordings, where the speaker activity is obtained from a diarization system. We show that this simple yet practical approach can successfully extract speakers after diarization, which results in improved ASR performance, especially in high overlapping conditions, with a relative word error rate reduction of up to 25%.

Hari Krishna Vydana, Martin Karafi'at, Katerina Zmolikova et al.

Conventional spoken language translation (SLT) systems are pipeline based systems, where we have an Automatic Speech Recognition (ASR) system to convert the modality of source from speech to text and a Machine Translation (MT) systems to translate source text to text in target language. Recent progress in the sequence-sequence architectures have reduced the performance gap between the pipeline based SLT systems (cascaded ASR-MT) and End-to-End approaches. Though End-to-End and cascaded ASR-MT systems are reaching to the comparable levels of performances, we can see a large performance gap using the ASR hypothesis and oracle text w.r.t MT models. This performance gap indicates that the MT systems are prone to large performance degradation due to noisy ASR hypothesis as opposed to oracle text transcript. In this work this degradation in the performance is reduced by creating an end to-end differentiable pipeline between the ASR and MT systems. In this work, we train SLT systems with ASR objective as an auxiliary loss and both the networks are connected through the neural hidden representations. This train ing would have an End-to-End differentiable path w.r.t to the final objective function as well as utilize the ASR objective for better performance of the SLT systems. This architecture has improved from BLEU from 36.8 to 44.5. Due to the Multi-task training the model also generates the ASR hypothesis which are used by a pre-trained MT model. Combining the proposed systems with the MT model has increased the BLEU score by 1. All the experiments are reported on English-Portuguese speech translation task using How2 corpus. The final BLEU score is on-par with the best speech translation system on How2 dataset with no additional training data and language model and much less parameters.

Marc Delcroix, Tsubasa Ochiai, Katerina Zmolikova et al.

Target speech extraction, which extracts a single target source in a mixture given clues about the target speaker, has attracted increasing attention. We have recently proposed SpeakerBeam, which exploits an adaptation utterance of the target speaker to extract his/her voice characteristics that are then used to guide a neural network towards extracting speech of that speaker. SpeakerBeam presents a practical alternative to speech separation as it enables tracking speech of a target speaker across utterances, and achieves promising speech extraction performance. However, it sometimes fails when speakers have similar voice characteristics, such as in same-gender mixtures, because it is difficult to discriminate the target speaker from the interfering speakers. In this paper, we investigate strategies for improving the speaker discrimination capability of SpeakerBeam. First, we propose a time-domain implementation of SpeakerBeam similar to that proposed for a time-domain audio separation network (TasNet), which has achieved state-of-the-art performance for speech separation. Besides, we investigate (1) the use of spatial features to better discriminate speakers when microphone array recordings are available, (2) adding an auxiliary speaker identification loss for helping to learn more discriminative voice characteristics. We show experimentally that these strategies greatly improve speech extraction performance, especially for same-gender mixtures, and outperform TasNet in terms of target speech extraction.