High-Fidelity Speech Synthesis with Minimal Supervision: All Using Diffusion Models

Text-to-speech (TTS) methods have shown promising results in voice cloning, but they require a large number of labeled text-speech pairs. Minimally-supervised speech synthesis decouples TTS by combining two types of discrete speech representations(semantic \& acoustic) and using two sequence-to-sequence tasks to enable training with minimal supervision. However, existing methods suffer from information redundancy and dimension explosion in semantic representation, and high-frequency waveform distortion in discrete acoustic representation. Autoregressive frameworks exhibit typical instability and uncontrollability issues. And non-autoregressive frameworks suffer from prosodic averaging caused by duration prediction models. To address these issues, we propose a minimally-supervised high-fidelity speech synthesis method, where all modules are constructed based on the diffusion models. The non-autoregressive framework enhances controllability, and the duration diffusion model enables diversified prosodic expression. Contrastive Token-Acoustic Pretraining (CTAP) is used as an intermediate semantic representation to solve the problems of information redundancy and dimension explosion in existing semantic coding methods. Mel-spectrogram is used as the acoustic representation. Both semantic and acoustic representations are predicted by continuous variable regression tasks to solve the problem of high-frequency fine-grained waveform distortion. Experimental results show that our proposed method outperforms the baseline method. We provide audio samples on our website.

Learning Speech Representation From Contrastive Token-Acoustic Pretraining

For fine-grained generation and recognition tasks such as minimally-supervised text-to-speech (TTS), voice conversion (VC), and automatic speech recognition (ASR), the intermediate representations extracted from speech should serve as a "bridge" between text and acoustic information, containing information from both modalities. The semantic content is emphasized, while the paralinguistic information such as speaker identity and acoustic details should be de-emphasized. However, existing methods for extracting fine-grained intermediate representations from speech suffer from issues of excessive redundancy and dimension explosion. Contrastive learning is a good method for modeling intermediate representations from two modalities. However, existing contrastive learning methods in the audio field focus on extracting global descriptive information for downstream audio classification tasks, making them unsuitable for TTS, VC, and ASR tasks. To address these issues, we propose a method named "Contrastive Token-Acoustic Pretraining (CTAP)", which uses two encoders to bring phoneme and speech into a joint multimodal space, learning how to connect phoneme and speech at the frame level. The CTAP model is trained on 210k speech and phoneme text pairs, achieving minimally-supervised TTS, VC, and ASR. The proposed CTAP method offers a promising solution for fine-grained generation and recognition downstream tasks in speech processing.

Minimally-Supervised Speech Synthesis with Conditional Diffusion Model and Language Model: A Comparative Study of Semantic Coding

Recently, there has been a growing interest in text-to-speech (TTS) methods that can be trained with minimal supervision by combining two types of discrete speech representations and using two sequence-to-sequence tasks to decouple TTS. To address the challenges associated with high dimensionality and waveform distortion in discrete representations, we propose Diff-LM-Speech, which models semantic embeddings into mel-spectrogram based on diffusion models and introduces a prompt encoder structure based on variational autoencoders and prosody bottlenecks to improve prompt representation capabilities. Autoregressive language models often suffer from missing and repeated words, while non-autoregressive frameworks face expression averaging problems due to duration prediction models. To address these issues, we propose Tetra-Diff-Speech, which designs a duration diffusion model to achieve diverse prosodic expressions. While we expect the information content of semantic coding to be between that of text and acoustic coding, existing models extract semantic coding with a lot of redundant information and dimensionality explosion. To verify that semantic coding is not necessary, we propose Tri-Diff-Speech. Experimental results show that our proposed methods outperform baseline methods. We provide a website with audio samples.

Rethinking the visual cues in audio-visual speaker extraction

The Audio-Visual Speaker Extraction (AVSE) algorithm employs parallel video recording to leverage two visual cues, namely speaker identity and synchronization, to enhance performance compared to audio-only algorithms. However, the visual front-end in AVSE is often derived from a pre-trained model or end-to-end trained, making it unclear which visual cue contributes more to the speaker extraction performance. This raises the question of how to better utilize visual cues. To address this issue, we propose two training strategies that decouple the learning of the two visual cues. Our experimental results demonstrate that both visual cues are useful, with the synchronization cue having a higher impact. We introduce a more explainable model, the Decoupled Audio-Visual Speaker Extraction (DAVSE) model, which leverages both visual cues.

speech and noise dual-stream spectrogram refine network with speech distortion loss for robust speech recognition

In recent years, the joint training of speech enhancement front-end and automatic speech recognition (ASR) back-end has been widely used to improve the robustness of ASR systems. Traditional joint training methods only use enhanced speech as input for the backend. However, it is difficult for speech enhancement systems to directly separate speech from input due to the diverse types of noise with different intensities. Furthermore, speech distortion and residual noise are often observed in enhanced speech, and the distortion of speech and noise is different. Most existing methods focus on fusing enhanced and noisy features to address this issue. In this paper, we propose a dual-stream spectrogram refine network to simultaneously refine the speech and noise and decouple the noise from the noisy input. Our proposed method can achieve better performance with a relative 8.6% CER reduction.

Locate and Beamform: Two-dimensional Locating All-neural Beamformer for Multi-channel Speech Separation

Recently, stunning improvements on multi-channel speech separation have been achieved by neural beamformers when direction information is available. However, most of them neglect to utilize speaker's 2-dimensional (2D) location cues contained in mixture signal, which limits the performance when two sources come from close directions. In this paper, we propose an end-to-end beamforming network for 2D location guided speech separation merely given mixture signal. It first estimates discriminable direction and 2D location cues, which imply directions the sources come from in multi views of microphones and their 2D coordinates. These cues are then integrated into location-aware neural beamformer, thus allowing accurate reconstruction of two sources' speech signals. Experiments show that our proposed model not only achieves a comprehensive decent improvement compared to baseline systems, but avoids inferior performance on spatial overlapping cases.

Time-domain Speech Enhancement Assisted by Multi-resolution Frequency Encoder and Decoder

Time-domain speech enhancement (SE) has recently been intensively investigated. Among recent works, DEMUCS introduces multi-resolution STFT loss to enhance performance. However, some resolutions used for STFT contain non-stationary signals, and it is challenging to learn multi-resolution frequency losses simultaneously with only one output. For better use of multi-resolution frequency information, we supplement multiple spectrograms in different frame lengths into the time-domain encoders. They extract stationary frequency information in both narrowband and wideband. We also adopt multiple decoder outputs, each of which computes its corresponding resolution frequency loss. Experimental results show that (1) it is more effective to fuse stationary frequency features than non-stationary features in the encoder, and (2) the multiple outputs consistent with the frequency loss improve performance. Experiments on the Voice-Bank dataset show that the proposed method obtained a 0.14 PESQ improvement.

Cross-modal Audio-visual Co-learning for Text-independent Speaker Verification

Visual speech (i.e., lip motion) is highly related to auditory speech due to the co-occurrence and synchronization in speech production. This paper investigates this correlation and proposes a cross-modal speech co-learning paradigm. The primary motivation of our cross-modal co-learning method is modeling one modality aided by exploiting knowledge from another modality. Specifically, two cross-modal boosters are introduced based on an audio-visual pseudo-siamese structure to learn the modality-transformed correlation. Inside each booster, a max-feature-map embedded Transformer variant is proposed for modality alignment and enhanced feature generation. The network is co-learned both from scratch and with pretrained models. Experimental results on the LRSLip3, GridLip, LomGridLip, and VoxLip datasets demonstrate that our proposed method achieves 60% and 20% average relative performance improvement over independently trained audio-only/visual-only and baseline fusion systems, respectively.

MIMO-DBnet: Multi-channel Input and Multiple Outputs DOA-aware Beamforming Network for Speech Separation

Recently, many deep learning based beamformers have been proposed for multi-channel speech separation. Nevertheless, most of them rely on extra cues known in advance, such as speaker feature, face image or directional information. In this paper, we propose an end-to-end beamforming network for direction guided speech separation given merely the mixture signal, namely MIMO-DBnet. Specifically, we design a multi-channel input and multiple outputs architecture to predict the direction-of-arrival based embeddings and beamforming weights for each source. The precisely estimated directional embedding provides quite effective spatial discrimination guidance for the neural beamformer to offset the effect of phase wrapping, thus allowing more accurate reconstruction of two sources' speech signals. Experiments show that our proposed MIMO-DBnet not only achieves a comprehensive decent improvement compared to baseline systems, but also maintain the performance on high frequency bands when phase wrapping occurs.

Monolingual Recognizers Fusion for Code-switching Speech Recognition

The bi-encoder structure has been intensively investigated in code-switching (CS) automatic speech recognition (ASR). However, most existing methods require the structures of two monolingual ASR models (MAMs) should be the same and only use the encoder of MAMs. This leads to the problem that pre-trained MAMs cannot be timely and fully used for CS ASR. In this paper, we propose a monolingual recognizers fusion method for CS ASR. It has two stages: the speech awareness (SA) stage and the language fusion (LF) stage. In the SA stage, acoustic features are mapped to two language-specific predictions by two independent MAMs. To keep the MAMs focused on their own language, we further extend the language-aware training strategy for the MAMs. In the LF stage, the BELM fuses two language-specific predictions to get the final prediction. Moreover, we propose a text simulation strategy to simplify the training process of the BELM and reduce reliance on CS data. Experiments on a Mandarin-English corpus show the efficiency of the proposed method. The mix error rate is significantly reduced on the test set after using open-source pre-trained MAMs.