Dual learning is a paradigm for semi-supervised machine learning that seeks to leverage unsupervised data by solving two opposite tasks at once. In this scheme, each model is used to generate pseudo-labels for unlabeled examples that are used to train the other model. Dual learning has seen some use in speech processing by pairing ASR and TTS as dual tasks. However, these results mostly address only the case of using unpaired examples to compensate for very small supervised datasets, and mostly on large, non-streaming models. Dual learning has not yet been proven effective for using unsupervised data to improve realistic on-device streaming models that are already trained on large supervised corpora. We provide this missing piece though an analysis of an on-device-sized streaming conformer trained on the entirety of Librispeech, showing relative WER improvements of 10.7%/5.2% without an LM and 11.7%/16.4% with an LM.
Neural speaker embeddings encode the speaker's speech characteristics through a DNN model and are prevalent for speaker verification tasks. However, few studies have investigated the usage of neural speaker embeddings for an ASR system. In this work, we present our efforts w.r.t integrating neural speaker embeddings into a conformer based hybrid HMM ASR system. For ASR, our improved embedding extraction pipeline in combination with the Weighted-Simple-Add integration method results in x-vector and c-vector reaching on par performance with i-vectors. We further compare and analyze different speaker embeddings. We present our acoustic model improvements obtained by switching from newbob learning rate schedule to one cycle learning schedule resulting in a ~3% relative WER reduction on Switchboard, additionally reducing the overall training time by 17%. By further adding neural speaker embeddings, we gain additional ~3% relative WER improvement on Hub5'00. Our best Conformer-based hybrid ASR system with speaker embeddings achieves 9.0% WER on Hub5'00 and Hub5'01 with training on SWB 300h.
The performances of automatic speech recognition (ASR) systems degrade drastically under noisy conditions. Explicit distortion modelling (EDM), as a feature compensation step, is able to enhance ASR systems under such conditions by simulating the in-domain noisy speeches from the clean counterparts. Yet, existing distortion models are either non-trainable or unexplainable and often lack controllability and generalization ability. In this paper, we propose a fully explainable and controllable model: DENT-DDSP to achieve EDM. DENT-DDSP utilizes novel differentiable digital signal processing (DDSP) components and requires only 10 seconds of training data to achieve high fidelity. The experiment shows that the simulated noisy data from DENT-DDSP achieves the highest simulation fidelity compared to other baseline models in terms of multi-scale spectral loss (MSSL). Moreover, to validate whether the data simulated by DENT-DDSP are able to replace the scarce in-domain noisy data in the noise-robust ASR tasks, several downstream ASR models with the same architecture are trained using the simulated data and the real data. The experiment shows that the model trained with the simulated noisy data from DENT-DDSP achieves similar performances to the benchmark with a 2.7\% difference in terms of word error rate (WER). The code of the model is released online.
The optimization of a wavelet-based algorithm to improve speech intelligibility is reported. The discrete-time speech signal is split into frequency sub-bands via a multi-level discrete wavelet transform. Various gains are applied to the sub-band signals before they are recombined to form a modified version of the speech. The sub-band gains are adjusted while keeping the overall signal energy unchanged, and the speech intelligibility under various background interference and simulated hearing loss conditions is enhanced and evaluated objectively and quantitatively using Google Speech-to-Text transcription. For English and Chinese noise-free speech, overall intelligibility is improved, and the transcription accuracy can be increased by as much as 80 percentage points by reallocating the spectral energy toward the mid-frequency sub-bands, effectively increasing the consonant-vowel intensity ratio. This is reasonable since the consonants are relatively weak and of short duration, which are therefore the most likely to become indistinguishable in the presence of background noise or high-frequency hearing impairment. For speech already corrupted by noise, improving intelligibility is challenging but still realizable. The proposed algorithm is implementable for real-time signal processing and comparatively simpler than previous algorithms. Potential applications include speech enhancement, hearing aids, machine listening, and a better understanding of speech intelligibility.
Talking head generation is to synthesize a lip-synchronized talking head video by inputting an arbitrary face image and corresponding audio clips. Existing methods ignore not only the interaction and relationship of cross-modal information, but also the local driving information of the mouth muscles. In this study, we propose a novel generative framework that contains a dilated non-causal temporal convolutional self-attention network as a multimodal fusion module to promote the relationship learning of cross-modal features. In addition, our proposed method uses both audio- and speech-related facial action units (AUs) as driving information. Speech-related AU information can guide mouth movements more accurately. Because speech is highly correlated with speech-related AUs, we propose an audio-to-AU module to predict speech-related AU information. We utilize pre-trained AU classifier to ensure that the generated images contain correct AU information. We verify the effectiveness of the proposed model on the GRID and TCD-TIMIT datasets. An ablation study is also conducted to verify the contribution of each component. The results of quantitative and qualitative experiments demonstrate that our method outperforms existing methods in terms of both image quality and lip-sync accuracy.
While the performance of offline neural speech separation systems has been greatly advanced by the recent development of novel neural network architectures, there is typically an inevitable performance gap between the systems and their online variants. In this paper, we investigate how RNN-based offline neural speech separation systems can be changed into their online counterparts while mitigating the performance degradation. We decompose or reorganize the forward and backward RNN layers in a bidirectional RNN layer to form an online path and an offline path, which enables the model to perform both online and offline processing with a same set of model parameters. We further introduce two training strategies for improving the online model via either a pretrained offline model or a multitask training objective. Experiment results show that compared to the online models that are trained from scratch, the proposed layer decomposition and reorganization schemes and training strategies can effectively mitigate the performance gap between two RNN-based offline separation models and their online variants.
Parallel text-to-speech (TTS) models have recently enabled fast and highly-natural speech synthesis. However, such models typically require external alignment models, which are not necessarily optimized for the decoder as they are not jointly trained. In this paper, we propose a differentiable duration method for learning monotonic alignments between input and output sequences. Our method is based on a soft-duration mechanism that optimizes a stochastic process in expectation. Using this differentiable duration method, a direct text to waveform TTS model is introduced to produce raw audio as output instead of performing neural vocoding. Our model learns to perform high-fidelity speech synthesis through a combination of adversarial training and matching the total ground-truth duration. Experimental results show that our model obtains competitive results while enjoying a much simpler training pipeline. Audio samples are available online.
In this paper, we discuss an important aspect of speech privacy: protecting spoken content. New capabilities from the field of machine learning provide a unique and timely opportunity to revisit speech content protection. There are many different applications of content privacy, even though this area has been under-explored in speech technology research. This paper presents several scenarios that indicate a need for speech content privacy even as the specific techniques to achieve content privacy may necessarily vary. Our discussion includes several different types of content privacy including recoverable and non-recoverable content. Finally, we introduce evaluation strategies as well as describe some of the difficulties that may be encountered.
Voiced speech signals such as continuous speech are known to have acoustic features such as pitch(F0), and formant frequencies(F1, F2, F3) which can be used for gender classification. However, gender classification studies using non-speech signals such as vocal breath sounds have not been explored as they lack typical gender-specific acoustic features. In this work, we explore whether vocal breath sounds encode gender information and if so, to what extent it can be used for automatic gender classification. In this study, we explore the use of data-driven and knowledge-based features from vocal breath sounds as well as the classifier complexity for gender classification. We also explore the importance of the location and duration of breath signal segments to be used for automatic classification. Experiments with 54.23 minutes of male and 51.83 minutes of female breath sounds reveal that knowledge-based features, namely MFCC statistics, with low-complexity classifier perform comparably to the data-driven features with classifiers of higher complexity. Breath segments with an average duration of 3 seconds are found to be the best choice irrespective of the location which avoids the need for breath cycle boundary annotation.
Stigma toward people who use substances (PWUS) is a leading barrier to seeking treatment. Further, those in treatment are more likely to drop out if they experience higher levels of stigmatization. While related concepts of hate speech and toxicity, including those targeted toward vulnerable populations, have been the focus of automatic content moderation research, stigma and, in particular, people who use substances have not. This paper explores stigma toward PWUS using a data set of roughly 5,000 public Reddit posts. We performed a crowd-sourced annotation task where workers are asked to annotate each post for the presence of stigma toward PWUS and answer a series of questions related to their experiences with substance use. Results show that workers who use substances or know someone with a substance use disorder are more likely to rate a post as stigmatizing. Building on this, we use a supervised machine learning framework that centers workers with lived substance use experience to label each Reddit post as stigmatizing. Modeling person-level demographics in addition to comment-level language results in a classification accuracy (as measured by AUC) of 0.69 -- a 17% increase over modeling language alone. Finally, we explore the linguist cues which distinguish stigmatizing content: PWUS substances and those who don't agree that language around othering ("people", "they") and terms like "addict" are stigmatizing, while PWUS (as opposed to those who do not) find discussions around specific substances more stigmatizing. Our findings offer insights into the nature of perceived stigma in substance use. Additionally, these results further establish the subjective nature of such machine learning tasks, highlighting the need for understanding their social contexts.