Electrical Engineering And Systems Science

Transformers are powerful neural architectures that allow integrating different modalities using attention mechanisms. In this paper, we leverage the neural transformer architectures for multi-channel speech recognition systems, where the spectral and spatial information collected from different microphones are integrated using attention layers. Our multi-channel transformer network mainly consists of three parts: channel-wise self attention layers (CSA), cross-channel attention layers (CCA), and multi-channel encoder-decoder attention layers (EDA). The CSA and CCA layers encode the contextual relationship within and between channels and across time, respectively. The channel-attended outputs from CSA and CCA are then fed into the EDA layers to help decode the next token given the preceding ones. The experiments show that in a far-field in-house dataset, our method outperforms the baseline single-channel transformer, as well as the super-directive and neural beamformers cascaded with the transformers.

A key desiderata for inclusive and accessible speech recognition technology is ensuring its robust performance to children's speech. Notably, this includes the rapidly advancing neural network based end-to-end speech recognition systems. Children speech recognition is more challenging due to the larger intra-inter speaker variability in terms of acoustic and linguistic characteristics compared to adult speech. Furthermore, the lack of adequate and appropriate children speech resources adds to the challenge of designing robust end-to-end neural architectures. This study provides a critical assessment of automatic children speech recognition through an empirical study of contemporary state-of-the-art end-to-end speech recognition systems. Insights are provided on the aspects of training data requirements, adaptation on children data, and the effect of children age, utterance lengths, different architectures and loss functions for end-to-end systems and role of language models on the speech recognition performance.

Voice activity detection (VAD) is an essential pre-processing step for tasks such as automatic speech recognition (ASR) and speaker recognition. A basic goal is to remove silent segments within an audio, while a more general VAD system could remove all the irrelevant segments such as noise and even unwanted speech from non-target speakers. We define the task, which only detects the speech from the target speaker, as speaker-dependent voice activity detection (SDVAD). This task is quite common in real applications and usually implemented by performing speaker verification (SV) on audio segments extracted from VAD. In this paper, we propose an end-to-end neural network based approach to address this problem, which explicitly takes the speaker identity into the modeling process. Moreover, inference can be performed in an online fashion, which leads to low system latency. Experiments are carried out on a conversational telephone dataset generated from the Switchboard corpus. Results show that our proposed online approach achieves significantly better performance than the usual VAD/SV system in terms of both frame accuracy and F-score. We also used our previously proposed segment-level metric for a more comprehensive analysis.

Disfluency detection is usually an intermediate step between an automatic speech recognition (ASR) system and a downstream task. By contrast, this paper aims to investigate the task of end-to-end speech recognition and disfluency removal. We specifically explore whether it is possible to train an ASR model to directly map disfluent speech into fluent transcripts, without relying on a separate disfluency detection model. We show that end-to-end models do learn to directly generate fluent transcripts; however, their performance is slightly worse than a baseline pipeline approach consisting of an ASR system and a disfluency detection model. We also propose two new metrics that can be used for evaluating integrated ASR and disfluency models. The findings of this paper can serve as a benchmark for further research on the task of end-to-end speech recognition and disfluency removal in the future.

Generalized end-to-end (GE2E) model is widely used in speaker verification (SV) fields due to its expandability and generality regardless of specific languages. However, the long-short term memory (LSTM) based on GE2E has two limitations: First, the embedding of GE2E suffers from vanishing gradient, which leads to performance degradation for very long input sequences. Secondly, utterances are not represented as a properly fixed dimensional vector. In this paper, to overcome issues mentioned above, we propose a novel framework for SV, end-to-end trainable self-attentive shallow network (SASN), incorporating a time-delay neural network (TDNN) and a self-attentive pooling mechanism based on the self-attentive x-vector system during an utterance embedding phase. We demonstrate that the proposed model is highly efficient, and provides more accurate speaker verification than GE2E. For VCTK dataset, with just less than half the size of GE2E, the proposed model showed significant performance improvement over GE2E of about 63%, 67%, and 85% in EER (Equal error rate), DCF (Detection cost function), and AUC (Area under the curve), respectively. Notably, when the input length becomes longer, the DCF score improvement of the proposed model is about 17 times greater than that of GE2E.

Automatic speech recognition (ASR) in multimedia content is one of the promising applications, but speech data in this kind of content are frequently mixed with background music, which is harmful for the performance of ASR. In this study, we propose a method for improving ASR with background music based on time-domain source separation. We utilize Conv-TasNet as a separation network, which has achieved state-of-the-art performance for multi-speaker source separation, to extract the speech signal from a speech-music mixture in the waveform domain. We also propose joint fine-tuning of a pre-trained Conv-TasNet front-end with an attention-based ASR back-end using both separation and ASR objectives. We evaluated our method through ASR experiments using speech data mixed with background music from a wide variety of Japanese animations. We show that time-domain speech-music separation drastically improves ASR performance of the back-end model trained with mixture data, and the joint optimization yielded a further significant WER reduction. The time-domain separation method outperformed a frequency-domain separation method, which reuses the phase information of the input mixture signal, both in simple cascading and joint training settings. We also demonstrate that our method works robustly for music interference from classical, jazz and popular genres.

Albeit recent progress in speaker verification generates powerful models, malicious attacks in the form of spoofed speech, are generally not coped with. Recent results in ASVSpoof2015 and BTAS2016 challenges indicate that spoof-aware features are a possible solution to this problem. Most successful methods in both challenges focus on spoof-aware features, rather than focusing on a powerful classifier. In this paper we present a novel raw waveform based deep model for spoofing detection, which jointly acts as a feature extractor and classifier, thus allowing it to directly classify speech signals. This approach can be considered as an end-to-end classifier, which removes the need for any pre- or post-processing on the data, making training and evaluation a streamlined process, consuming less time than other neural-network based approaches. The experiments on the BTAS2016 dataset show that the system performance is significantly improved by the proposed raw waveform convolutional long short term neural network (CLDNN), from the previous best published 1.26\% half total error rate (HTER) to the current 0.82\% HTER. Moreover it shows that the proposed system also performs well under the unknown (RE-PH2-PH3,RE-LPPH2-PH3) conditions.

Audio codecs based on discretized neural autoencoders have recently been developed and shown to provide significantly higher compression levels for comparable quality speech output. However, these models are tightly coupled with speech content, and produce unintended outputs in noisy conditions. Based on VQ-VAE autoencoders with WaveRNN decoders, we develop compressor-enhancer encoders and accompanying decoders, and show that they operate well in noisy conditions. We also observe that a compressor-enhancer model performs better on clean speech inputs than a compressor model trained only on clean speech.

We study the individuality of the human voice with respect to a widely used feature representation of speech utterances, namely, the i-vector model. As a first step toward this goal, we compare and contrast uniqueness measures proposed for different biometric modalities. Then, we introduce a new uniqueness measure that evaluates the entropy of i-vectors while taking into account speaker level variations. Our measure operates in the discrete feature space and relies on accurate estimation of the distribution of i-vectors. Therefore, i-vectors are quantized while ensuring that both the quantized and original representations yield similar speaker verification performance. Uniqueness estimates are obtained from two newly generated datasets and the public VoxCeleb dataset. The first custom dataset contains more than one and a half million speech samples of 20,741 speakers obtained from TEDx Talks videos. The second one includes over twenty one thousand speech samples from 1,595 actors that are extracted from movie dialogues. Using this data, we analyzed how several factors, such as the number of speakers, number of samples per speaker, sample durations, and diversity of utterances affect uniqueness estimates. Most notably, we determine that the discretization of i-vectors does not cause a reduction in speaker recognition performance. Our results show that the degree of distinctiveness offered by i-vector-based representation may reach 43-70 bits considering 5-second long speech samples; however, under less constrained variations in speech, uniqueness estimates are found to reduce by around 30 bits. We also find that doubling the sample duration increases the distinctiveness of the i-vector representation by around 20 bits.

In this paper we present a privacy-aware method for estimating source-dominated microphone clusters in the context of acoustic sensor networks (ASNs). The approach is based on clustered federated learning which we adapt to unsupervised scenarios by employing a light-weight autoencoder model. The model is further optimized for training on very scarce data. In order to best harness the benefits of clustered microphone nodes in ASN applications, a method for the computation of cluster membership values is introduced. We validate the performance of the proposed approach using clustering-based measures and a network-wide classification task.