Yujin Wang

Ziyang Ma, Zhisheng Zheng, Changli Tang et al.

In this paper, we provide a new perspective on self-supervised speech models from how the training targets are obtained. We generalize the targets extractor into Offline Targets Extractor (Off-TE) and Online Targets Extractor (On-TE). Based on this, we propose a new multi-tasking learning framework for self-supervised learning, MT4SSL, which stands for Boosting Self-Supervised Speech Representation Learning by Integrating Multiple Targets. MT4SSL uses the K-means algorithm as an Off-TE and a teacher network without gradients as an On-TE, respectively. Our model outperforms previous SSL methods by nontrivial margins on the LibriSpeech benchmark, and is comparable to or even better than the best-performing models with fewer data. Furthermore, we find that using both Off-TE and On-TE results in better convergence in the pre-training phase. With both effectiveness and efficiency, we think doing multi-task learning on self-supervised speech models from our perspective is a promising trend.

Xie Chen, Ziyang Ma, Changli Tang et al.

Recent years have witnessed a boom in self-supervised learning (SSL) in various areas including speech processing. Speech based SSL models present promising performance in a range of speech related tasks. However, the training of SSL models is computationally expensive and a common practice is to fine-tune a released SSL model on the specific task. It is essential to use consistent front-end input during pre-training and fine-tuning. This consistency may introduce potential issues when the optimal front-end is not the same as that used in pre-training. In this paper, we propose a simple but effective front-end adapter to address this front-end discrepancy. By minimizing the distance between the outputs of different front-ends, the filterbank feature (Fbank) can be compatible with SSL models which are pre-trained with waveform. The experiment results demonstrate the effectiveness of our proposed front-end adapter on several popular SSL models for the speech recognition task.

Yujin Wang, Changli Tang, Ziyang Ma et al.

Recent years have witnessed great strides in self-supervised learning (SSL) on the speech processing. The SSL model is normally pre-trained on a great variety of unlabelled data and a large model size is preferred to increase the modeling capacity. However, this might limit its potential applications due to the expensive computation and memory costs introduced by the oversize model. Miniaturization for SSL models has become an important research direction of practical value. To this end, we explore the effective distillation of HuBERT-based SSL models for automatic speech recognition (ASR). First, in order to establish a strong baseline, a comprehensive study on different student model structures is conducted. On top of this, as a supplement to the regression loss widely adopted in previous works, a discriminative loss is introduced for HuBERT to enhance the distillation performance, especially in low-resource scenarios. In addition, we design a simple and effective algorithm to distill the front-end input from waveform to Fbank feature, resulting in 17% parameter reduction and doubling inference speed, at marginal performance degradation.

Yujin Wang, Lingen Li, Tianfan Xue et al.

Image denoising is a fundamental and challenging task in the field of computer vision. Most supervised denoising methods learn to reconstruct clean images from noisy inputs, which have intrinsic spectral bias and tend to produce over-smoothed and blurry images. Recently, researchers have explored diffusion models to generate high-frequency details in image restoration tasks, but these models do not guarantee that the generated texture aligns with real images, leading to undesirable artifacts. To address the trade-off between visual appeal and fidelity of high-frequency details in denoising tasks, we propose a novel approach called the Reconstruct-and-Generate Diffusion Model (RnG). Our method leverages a reconstructive denoising network to recover the majority of the underlying clean signal, which serves as the initial estimation for subsequent steps to maintain fidelity. Additionally, it employs a diffusion algorithm to generate residual high-frequency details, thereby enhancing visual quality. We further introduce a two-stage training scheme to ensure effective collaboration between the reconstructive and generative modules of RnG. To reduce undesirable texture introduced by the diffusion model, we also propose an adaptive step controller that regulates the number of inverse steps applied by the diffusion model, allowing control over the level of high-frequency details added to each patch as well as saving the inference computational cost. Through our proposed RnG, we achieve a better balance between perception and distortion. We conducted extensive experiments on both synthetic and real denoising datasets, validating the superiority of the proposed approach.