Speaker-independent speech separation has achieved remarkable performance in recent years with the development of deep neural network (DNN). Various network architectures, from traditional convolutional neural network (CNN) and recurrent neural network (RNN) to advanced transformer, have been designed sophistically to improve separation performance. However, the state-of-the-art models usually suffer from several flaws related to the computation, such as large model size, huge memory consumption and computational complexity. To find the balance between the performance and computational efficiency and to further explore the modeling ability of traditional network structure, we combine RNN and a newly proposed variant of convolutional network to cope with speech separation problem. By embedding two RNNs into basic block of this variant with the help of dual-path strategy, the proposed network can effectively learn the local information and global dependency. Besides, a four-staged structure enables the separation procedure to be performed gradually at finer and finer scales as the feature dimension increases. The experimental results on various datasets have proven the effectiveness of the proposed method and shown that a trade-off between the separation performance and computational efficiency is well achieved.
When domain experts are needed to perform data annotation for complex machine-learning tasks, reducing annotation effort is crucial in order to cut down time and expenses. For cases when there are no annotations available, one approach is to utilize the structure of the feature space for clustering-based active learning (AL) methods. However, these methods are heavily dependent on how the samples are organized in the feature space and what distance metric is used. Unsupervised methods such as contrastive predictive coding (CPC) can potentially be used to learn organized feature spaces, but these methods typically create high-dimensional features which might be challenging for estimating data density. In this paper, we combine CPC and multiple dimensionality reduction methods in search of functioning practices for clustering-based AL. Our experiments for simulating speech emotion recognition system deployment show that both the local and global topology of the feature space can be successfully used for AL, and that CPC can be used to improve clustering-based AL performance over traditional signal features. Additionally, we observe that compressing data dimensionality does not harm AL performance substantially, and that 2-D feature representations achieved similar AL performance as higher-dimensional representations when the number of annotations is not very low.
Text-to-Speech (TTS) synthesis plays an important role in human-computer interaction. Currently, most TTS technologies focus on the naturalness of speech, namely,making the speeches sound like humans. However, the key tasks of the expression of emotion and the speaker identity are ignored, which limits the application scenarios of TTS synthesis technology. To make the synthesized speech more realistic and expand the application scenarios, we propose a multi-task anthropomorphic speech synthesis framework (MASS), which can synthesize speeches from text with specified emotion and speaker identity. The MASS framework consists of a base TTS module and two novel voice conversion modules: the emotional voice conversion module and the speaker voice conversion module. We propose deep emotion voice conversion model (DEVC) and deep speaker voice conversion model (DSVC) based on convolution residual networks. It solves the problem of feature loss during voice conversion. The model trainings are independent of parallel datasets, and are capable of many-to-many voice conversion. In the emotional voice conversion, speaker voice conversion experiments, as well as the multi-task speech synthesis experiments, experimental results show DEVC and DSVC convert speech effectively. The quantitative and qualitative evaluation results of multi-task speech synthesis experiments show MASS can effectively synthesis speech with specified text, emotion and speaker identity.
The high cost of data acquisition makes Automatic Speech Recognition (ASR) model training problematic for most existing languages, including languages that do not even have a written script, or for which the phone inventories remain unknown. Past works explored multilingual training, transfer learning, as well as zero-shot learning in order to build ASR systems for these low-resource languages. While it has been shown that the pooling of resources from multiple languages is helpful, we have not yet seen a successful application of an ASR model to a language unseen during training. A crucial step in the adaptation of ASR from seen to unseen languages is the creation of the phone inventory of the unseen language. The ultimate goal of our work is to build the phone inventory of a language unseen during training in an unsupervised way without any knowledge about the language. In this paper, we 1) investigate the influence of different factors (i.e., model architecture, phonotactic model, type of speech representation) on phone recognition in an unknown language; 2) provide an analysis of which phones transfer well across languages and which do not in order to understand the limitations of and areas for further improvement for automatic phone inventory creation; and 3) present different methods to build a phone inventory of an unseen language in an unsupervised way. To that end, we conducted mono-, multi-, and crosslingual experiments on a set of 13 phonetically diverse languages and several in-depth analyses. We found a number of universal phone tokens (IPA symbols) that are well-recognized cross-linguistically. Through a detailed analysis of results, we conclude that unique sounds, similar sounds, and tone languages remain a major challenge for phonetic inventory discovery.
Models that can handle a wide range of speakers and acoustic conditions are essential in speech emotion recognition (SER). Often, these models tend to show mixed results when presented with speakers or acoustic conditions that were not visible during training. This paper investigates the impact of cross-corpus data complementation and data augmentation on the performance of SER models in matched (test-set from same corpus) and mismatched (test-set from different corpus) conditions. Investigations using six emotional speech corpora that include single and multiple speakers as well as variations in emotion style (acted, elicited, natural) and recording conditions are presented. Observations show that, as expected, models trained on single corpora perform best in matched conditions while performance decreases between 10-40% in mismatched conditions, depending on corpus specific features. Models trained on mixed corpora can be more stable in mismatched contexts, and the performance reductions range from 1 to 8% when compared with single corpus models in matched conditions. Data augmentation yields additional gains up to 4% and seem to benefit mismatched conditions more than matched ones.
The increasing demand for learning English as a second language has led to a growing interest in methods for automatically assessing spoken language proficiency. Most approaches use hand-crafted features, but their efficacy relies on their particular underlying assumptions and they risk discarding potentially salient information about proficiency. Other approaches rely on transcriptions produced by ASR systems which may not provide a faithful rendition of a learner's utterance in specific scenarios (e.g., non-native children's spontaneous speech). Furthermore, transcriptions do not yield any information about relevant aspects such as intonation, rhythm or prosody. In this paper, we investigate the use of wav2vec 2.0 for assessing overall and individual aspects of proficiency on two small datasets, one of which is publicly available. We find that this approach significantly outperforms the BERT-based baseline system trained on ASR and manual transcriptions used for comparison.
Existing speech enhancement methods mainly separate speech from noises at the signal level or in the time-frequency domain. They seldom pay attention to the semantic information of a corrupted signal. In this paper, we aim to bridge this gap by extracting phoneme identities to help speech enhancement. Specifically, we propose a phoneme-based distribution regularization (PbDr) for speech enhancement, which incorporates frame-wise phoneme information into speech enhancement network in a conditional manner. As different phonemes always lead to different feature distributions in frequency, we propose to learn a parameter pair, i.e. scale and bias, through a phoneme classification vector to modulate the speech enhancement network. The modulation parameter pair includes not only frame-wise but also frequency-wise conditions, which effectively map features to phoneme-related distributions. In this way, we explicitly regularize speech enhancement features by recognition vectors. Experiments on public datasets demonstrate that the proposed PbDr module can not only boost the perceptual quality for speech enhancement but also the recognition accuracy of an ASR system on the enhanced speech. This PbDr module could be readily incorporated into other speech enhancement networks as well.
Adversarial attack approaches to speaker identification either need high computational cost or are not very effective, to our knowledge. To address this issue, in this paper, we propose a novel generation-network-based approach, called symmetric saliency-based encoder-decoder (SSED), to generate adversarial voice examples to speaker identification. It contains two novel components. First, it uses a novel saliency map decoder to learn the importance of speech samples to the decision of a targeted speaker identification system, so as to make the attacker focus on generating artificial noise to the important samples. It also proposes an angular loss function to push the speaker embedding far away from the source speaker. Our experimental results demonstrate that the proposed SSED yields the state-of-the-art performance, i.e. over 97% targeted attack success rate and a signal-to-noise level of over 39 dB on both the open-set and close-set speaker identification tasks, with a low computational cost.
We study the statistical properties of learning to defer (L2D) to multiple experts. In particular, we address the open problems of deriving a consistent surrogate loss, confidence calibration, and principled ensembling of experts. Firstly, we derive two consistent surrogates -- one based on a softmax parameterization, the other on a one-vs-all (OvA) parameterization -- that are analogous to the single expert losses proposed by Mozannar and Sontag (2020) and Verma and Nalisnick (2022), respectively. We then study the frameworks' ability to estimate P( m_j = y | x ), the probability that the jth expert will correctly predict the label for x. Theory shows the softmax-based loss causes mis-calibration to propagate between the estimates while the OvA-based loss does not (though in practice, we find there are trade offs). Lastly, we propose a conformal inference technique that chooses a subset of experts to query when the system defers. We perform empirical validation on tasks for galaxy, skin lesion, and hate speech classification.
Boundary information is critical for various Chinese language processing tasks, such as word segmentation, part-of-speech tagging, and named entity recognition. Previous studies usually resorted to the use of a high-quality external lexicon, where lexicon items can offer explicit boundary information. However, to ensure the quality of the lexicon, great human effort is always necessary, which has been generally ignored. In this work, we suggest unsupervised statistical boundary information instead, and propose an architecture to encode the information directly into pre-trained language models, resulting in Boundary-Aware BERT (BABERT). We apply BABERT for feature induction of Chinese sequence labeling tasks. Experimental results on ten benchmarks of Chinese sequence labeling demonstrate that BABERT can provide consistent improvements on all datasets. In addition, our method can complement previous supervised lexicon exploration, where further improvements can be achieved when integrated with external lexicon information.