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.
This paper describes the ESPnet Unsupervised ASR Open-source Toolkit (EURO), an end-to-end open-source toolkit for unsupervised automatic speech recognition (UASR). EURO adopts the state-of-the-art UASR learning method introduced by the Wav2vec-U, originally implemented at FAIRSEQ, which leverages self-supervised speech representations and adversarial training. In addition to wav2vec2, EURO extends the functionality and promotes reproducibility for UASR tasks by integrating S3PRL and k2, resulting in flexible frontends from 27 self-supervised models and various graph-based decoding strategies. EURO is implemented in ESPnet and follows its unified pipeline to provide UASR recipes with a complete setup. This improves the pipeline's efficiency and allows EURO to be easily applied to existing datasets in ESPnet. Extensive experiments on three mainstream self-supervised models demonstrate the toolkit's effectiveness and achieve state-of-the-art UASR performance on TIMIT and LibriSpeech datasets. EURO will be publicly available at https://github.com/espnet/espnet, aiming to promote this exciting and emerging research area based on UASR through open-source activity.
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.
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.
Training multilingual Neural Text-To-Speech (NTTS) models using only monolingual corpora has emerged as a popular way for building voice cloning based Polyglot NTTS systems. In order to train these models, it is essential to understand how the composition of the training corpora affects the quality of multilingual speech synthesis. In this context, it is common to hear questions such as "Would including more Spanish data help my Italian synthesis, given the closeness of both languages?". Unfortunately, we found existing literature on the topic lacking in completeness in this regard. In the present work, we conduct an extensive ablation study aimed at understanding how various factors of the training corpora, such as language family affiliation, gender composition, and the number of speakers, contribute to the quality of Polyglot synthesis. Our findings include the observation that female speaker data are preferred in most scenarios, and that it is not always beneficial to have more speakers from the target language variant in the training corpus. The findings herein are informative for the process of data procurement and corpora building.
This paper presents an optimized methodology to design and deploy Speech Enhancement (SE) algorithms based on Recurrent Neural Networks (RNNs) on a state-of-the-art MicroController Unit (MCU), with 1+8 general-purpose RISC-V cores. To achieve low-latency execution, we propose an optimized software pipeline interleaving parallel computation of LSTM or GRU recurrent blocks, featuring vectorized 8-bit integer (INT8) and 16-bit floating-point (FP16) compute units, with manually-managed memory transfers of model parameters. To ensure minimal accuracy degradation with respect to the full-precision models, we propose a novel FP16-INT8 Mixed-Precision Post-Training Quantization (PTQ) scheme that compresses the recurrent layers to 8-bit while the bit precision of remaining layers is kept to FP16. Experiments are conducted on multiple LSTM and GRU based SE models trained on the Valentini dataset, featuring up to 1.24M parameters. Thanks to the proposed approaches, we speed-up the computation by up to 4x with respect to the lossless FP16 baselines. Differently from a uniform 8-bit quantization that degrades the PESQ score by 0.3 on average, the Mixed-Precision PTQ scheme leads to a low-degradation of only 0.06, while achieving a 1.4-1.7x memory saving. Thanks to this compression, we cut the power cost of the external memory by fitting the large models on the limited on-chip non-volatile memory and we gain a MCU power saving of up to 2.5x by reducing the supply voltage from 0.8V to 0.65V while still matching the real-time constraints. Our design results 10x more energy efficient than state-of-the-art SE solutions deployed on single-core MCUs that make use of smaller models and quantization-aware training.
Because the performance of speech separation is excellent for speech in which two speakers completely overlap, research attention has been shifted to dealing with more realistic scenarios. However, domain mismatch between training/test situations due to factors, such as speaker, content, channel, and environment, remains a severe problem for speech separation. Speaker and environment mismatches have been studied in the existing literature. Nevertheless, there are few studies on speech content and channel mismatches. Moreover, the impacts of language and channel in these studies are mostly tangled. In this study, we create several datasets for various experiments. The results show that the impacts of different languages are small enough to be ignored compared to the impacts of different channels. In our experiments, training on data recorded by Android phones leads to the best generalizability. Moreover, we provide a new solution for channel mismatch by evaluating projection, where the channel similarity can be measured and used to effectively select additional training data to improve the performance of in-the-wild test data.
This paper described the PCG-AIID system for L3DAS22 challenge in Task 1: 3D speech enhancement in office reverberant environment. We proposed a two-stage framework to address multi-channel speech denoising and dereverberation. In the first stage, a multiple input and multiple output (MIMO) network is applied to remove background noise while maintaining the spatial characteristics of multi-channel signals. In the second stage, a multiple input and single output (MISO) network is applied to enhance the speech from desired direction and post-filtering. As a result, our system ranked 3rd place in ICASSP2022 L3DAS22 challenge and significantly outperforms the baseline system, while achieving 3.2% WER and 0.972 STOI on the blind test-set.
End-to-End deep learning has shown promising results for speech enhancement tasks, such as noise suppression, dereverberation, and speech separation. However, most state-of-the-art methods for echo cancellation are either classical DSP-based or hybrid DSP-ML algorithms. Components such as the delay estimator and adaptive linear filter are based on traditional signal processing concepts, and deep learning algorithms typically only serve to replace the non-linear residual echo suppressor. This paper introduces an end-to-end echo cancellation network with a streaming cross-attention alignment (SCA). Our proposed method can handle unaligned inputs without requiring external alignment and generate high-quality speech without echoes. At the same time, the end-to-end algorithm simplifies the current echo cancellation pipeline for time-variant echo path cases. We test our proposed method on the ICASSP2022 and Interspeech2021 Microsoft deep echo cancellation challenge evaluation dataset, where our method outperforms some of the other hybrid and end-to-end methods.
This paper describes a practical dual-process speech enhancement system that adapts environment-sensitive frame-online beamforming (front-end) with help from environment-free block-online source separation (back-end). To use minimum variance distortionless response (MVDR) beamforming, one may train a deep neural network (DNN) that estimates time-frequency masks used for computing the covariance matrices of sources (speech and noise). Backpropagation-based run-time adaptation of the DNN was proposed for dealing with the mismatched training-test conditions. Instead, one may try to directly estimate the source covariance matrices with a state-of-the-art blind source separation method called fast multichannel non-negative matrix factorization (FastMNMF). In practice, however, neither the DNN nor the FastMNMF can be updated in a frame-online manner due to its computationally-expensive iterative nature. Our DNN-free system leverages the posteriors of the latest source spectrograms given by block-online FastMNMF to derive the current source covariance matrices for frame-online beamforming. The evaluation shows that our frame-online system can quickly respond to scene changes caused by interfering speaker movements and outperformed an existing block-online system with DNN-based beamforming by 5.0 points in terms of the word error rate.