Lei He

Yichong Leng, Zehua Chen, Junliang Guo et al. · microsoft-research

Binaural audio plays a significant role in constructing immersive augmented and virtual realities. As it is expensive to record binaural audio from the real world, synthesizing them from mono audio has attracted increasing attention. This synthesis process involves not only the basic physical warping of the mono audio, but also room reverberations and head/ear related filtrations, which, however, are difficult to accurately simulate in traditional digital signal processing. In this paper, we formulate the synthesis process from a different perspective by decomposing the binaural audio into a common part that shared by the left and right channels as well as a specific part that differs in each channel. Accordingly, we propose BinauralGrad, a novel two-stage framework equipped with diffusion models to synthesize them respectively. Specifically, in the first stage, the common information of the binaural audio is generated with a single-channel diffusion model conditioned on the mono audio, based on which the binaural audio is generated by a two-channel diffusion model in the second stage. Combining this novel perspective of two-stage synthesis with advanced generative models (i.e., the diffusion models),the proposed BinauralGrad is able to generate accurate and high-fidelity binaural audio samples. Experiment results show that on a benchmark dataset, BinauralGrad outperforms the existing baselines by a large margin in terms of both object and subject evaluation metrics (Wave L2: 0.128 vs. 0.157, MOS: 3.80 vs. 3.61). The generated audio samples (https://speechresearch.github.io/binauralgrad) and code (https://github.com/microsoft/NeuralSpeech/tree/master/BinauralGrad) are available online.

Xu Tan, Jiawei Chen, Haohe Liu et al. · microsoft-research

Text to speech (TTS) has made rapid progress in both academia and industry in recent years. Some questions naturally arise that whether a TTS system can achieve human-level quality, how to define/judge that quality and how to achieve it. In this paper, we answer these questions by first defining the human-level quality based on the statistical significance of subjective measure and introducing appropriate guidelines to judge it, and then developing a TTS system called NaturalSpeech that achieves human-level quality on a benchmark dataset. Specifically, we leverage a variational autoencoder (VAE) for end-to-end text to waveform generation, with several key modules to enhance the capacity of the prior from text and reduce the complexity of the posterior from speech, including phoneme pre-training, differentiable duration modeling, bidirectional prior/posterior modeling, and a memory mechanism in VAE. Experiment evaluations on popular LJSpeech dataset show that our proposed NaturalSpeech achieves -0.01 CMOS (comparative mean opinion score) to human recordings at the sentence level, with Wilcoxon signed rank test at p-level p >> 0.05, which demonstrates no statistically significant difference from human recordings for the first time on this dataset.

Kai Shen, Zeqian Ju, Xu Tan et al. · microsoft-research

Scaling text-to-speech (TTS) to large-scale, multi-speaker, and in-the-wild datasets is important to capture the diversity in human speech such as speaker identities, prosodies, and styles (e.g., singing). Current large TTS systems usually quantize speech into discrete tokens and use language models to generate these tokens one by one, which suffer from unstable prosody, word skipping/repeating issue, and poor voice quality. In this paper, we develop NaturalSpeech 2, a TTS system that leverages a neural audio codec with residual vector quantizers to get the quantized latent vectors and uses a diffusion model to generate these latent vectors conditioned on text input. To enhance the zero-shot capability that is important to achieve diverse speech synthesis, we design a speech prompting mechanism to facilitate in-context learning in the diffusion model and the duration/pitch predictor. We scale NaturalSpeech 2 to large-scale datasets with 44K hours of speech and singing data and evaluate its voice quality on unseen speakers. NaturalSpeech 2 outperforms previous TTS systems by a large margin in terms of prosody/timbre similarity, robustness, and voice quality in a zero-shot setting, and performs novel zero-shot singing synthesis with only a speech prompt. Audio samples are available at https://speechresearch.github.io/naturalspeech2.

Yihan Wu, Junliang Guo, Xu Tan et al. · microsoft-research

Video dubbing aims to translate the original speech in a film or television program into the speech in a target language, which can be achieved with a cascaded system consisting of speech recognition, machine translation and speech synthesis. To ensure the translated speech to be well aligned with the corresponding video, the length/duration of the translated speech should be as close as possible to that of the original speech, which requires strict length control. Previous works usually control the number of words or characters generated by the machine translation model to be similar to the source sentence, without considering the isochronicity of speech as the speech duration of words/characters in different languages varies. In this paper, we propose a machine translation system tailored for the task of video dubbing, which directly considers the speech duration of each token in translation, to match the length of source and target speech. Specifically, we control the speech length of generated sentence by guiding the prediction of each word with the duration information, including the speech duration of itself as well as how much duration is left for the remaining words. We design experiments on four language directions (German -> English, Spanish -> English, Chinese <-> English), and the results show that the proposed method achieves better length control ability on the generated speech than baseline methods. To make up the lack of real-world datasets, we also construct a real-world test set collected from films to provide comprehensive evaluations on the video dubbing task.

Jie Gui, Xiaofeng Cong, Lei He et al.

On the one hand, the dehazing task is an illposedness problem, which means that no unique solution exists. On the other hand, the dehazing task should take into account the subjective factor, which is to give the user selectable dehazed images rather than a single result. Therefore, this paper proposes a multi-output dehazing network by introducing illumination controllable ability, called IC-Dehazing. The proposed IC-Dehazing can change the illumination intensity by adjusting the factor of the illumination controllable module, which is realized based on the interpretable Retinex theory. Moreover, the backbone dehazing network of IC-Dehazing consists of a Transformer with double decoders for high-quality image restoration. Further, the prior-based loss function and unsupervised training strategy enable IC-Dehazing to complete the parameter learning process without the need for paired data. To demonstrate the effectiveness of the proposed IC-Dehazing, quantitative and qualitative experiments are conducted on image dehazing, semantic segmentation, and object detection tasks. Code is available at https://github.com/Xiaofeng-life/ICDehazing.

Jiawei Yang, Hanbo Chen, Yu Zhao et al.

Whole slide image (WSI) classification often relies on deep weakly supervised multiple instance learning (MIL) methods to handle gigapixel resolution images and slide-level labels. Yet the decent performance of deep learning comes from harnessing massive datasets and diverse samples, urging the need for efficient training pipelines for scaling to large datasets and data augmentation techniques for diversifying samples. However, current MIL-based WSI classification pipelines are memory-expensive and computation-inefficient since they usually assemble tens of thousands of patches as bags for computation. On the other hand, despite their popularity in other tasks, data augmentations are unexplored for WSI MIL frameworks. To address them, we propose ReMix, a general and efficient framework for MIL based WSI classification. It comprises two steps: reduce and mix. First, it reduces the number of instances in WSI bags by substituting instances with instance prototypes, i.e., patch cluster centroids. Then, we propose a ``Mix-the-bag'' augmentation that contains four online, stochastic and flexible latent space augmentations. It brings diverse and reliable class-identity-preserving semantic changes in the latent space while enforcing semantic-perturbation invariance. We evaluate ReMix on two public datasets with two state-of-the-art MIL methods. In our experiments, consistent improvements in precision, accuracy, and recall have been achieved but with orders of magnitude reduced training time and memory consumption, demonstrating ReMix's effectiveness and efficiency. Code is available.

Yuancheng Wang, Zeqian Ju, Xu Tan et al.

Audio editing is applicable for various purposes, such as adding background sound effects, replacing a musical instrument, and repairing damaged audio. Recently, some diffusion-based methods achieved zero-shot audio editing by using a diffusion and denoising process conditioned on the text description of the output audio. However, these methods still have some problems: 1) they have not been trained on editing tasks and cannot ensure good editing effects; 2) they can erroneously modify audio segments that do not require editing; 3) they need a complete description of the output audio, which is not always available or necessary in practical scenarios. In this work, we propose AUDIT, an instruction-guided audio editing model based on latent diffusion models. Specifically, AUDIT has three main design features: 1) we construct triplet training data (instruction, input audio, output audio) for different audio editing tasks and train a diffusion model using instruction and input (to be edited) audio as conditions and generating output (edited) audio; 2) it can automatically learn to only modify segments that need to be edited by comparing the difference between the input and output audio; 3) it only needs edit instructions instead of full target audio descriptions as text input. AUDIT achieves state-of-the-art results in both objective and subjective metrics for several audio editing tasks (e.g., adding, dropping, replacement, inpainting, super-resolution). Demo samples are available at https://audit-demo.github.io/.

Zehua Chen, Yihan Wu, Yichong Leng et al.

Denoising Diffusion Probabilistic Models (DDPMs) are emerging in text-to-speech (TTS) synthesis because of their strong capability of generating high-fidelity samples. However, their iterative refinement process in high-dimensional data space results in slow inference speed, which restricts their application in real-time systems. Previous works have explored speeding up by minimizing the number of inference steps but at the cost of sample quality. In this work, to improve the inference speed for DDPM-based TTS model while achieving high sample quality, we propose ResGrad, a lightweight diffusion model which learns to refine the output spectrogram of an existing TTS model (e.g., FastSpeech 2) by predicting the residual between the model output and the corresponding ground-truth speech. ResGrad has several advantages: 1) Compare with other acceleration methods for DDPM which need to synthesize speech from scratch, ResGrad reduces the complexity of task by changing the generation target from ground-truth mel-spectrogram to the residual, resulting into a more lightweight model and thus a smaller real-time factor. 2) ResGrad is employed in the inference process of the existing TTS model in a plug-and-play way, without re-training this model. We verify ResGrad on the single-speaker dataset LJSpeech and two more challenging datasets with multiple speakers (LibriTTS) and high sampling rate (VCTK). Experimental results show that in comparison with other speed-up methods of DDPMs: 1) ResGrad achieves better sample quality with the same inference speed measured by real-time factor; 2) with similar speech quality, ResGrad synthesizes speech faster than baseline methods by more than 10 times. Audio samples are available at https://resgrad1.github.io/.

Yangyang Guo, Hao Wang, Lei He et al.

An efficient team is essential for the company to successfully complete new projects. To solve the team formation problem considering person-job matching (TFP-PJM), a 0-1 integer programming model is constructed, which considers both person-job matching and team members' willingness to communicate on team efficiency, with the person-job matching score calculated using intuitionistic fuzzy numbers. Then, a reinforcement learning-assisted genetic programming algorithm (RL-GP) is proposed to enhance the quality of solutions. The RL-GP adopts the ensemble population strategies. Before the population evolution at each generation, the agent selects one from four population search modes according to the information obtained, thus realizing a sound balance of exploration and exploitation. In addition, surrogate models are used in the algorithm to evaluate the formation plans generated by individuals, which speeds up the algorithm learning process. Afterward, a series of comparison experiments are conducted to verify the overall performance of RL-GP and the effectiveness of the improved strategies within the algorithm. The hyper-heuristic rules obtained through efficient learning can be utilized as decision-making aids when forming project teams. This study reveals the advantages of reinforcement learning methods, ensemble strategies, and the surrogate model applied to the GP framework. The diversity and intelligent selection of search patterns along with fast adaptation evaluation, are distinct features that enable RL-GP to be deployed in real-world enterprise environments.

Weinan Song, Yaxuan Zhu, Lei He et al.

This paper studies a novel energy-based cooperative learning framework for multi-domain image-to-image translation. The framework consists of four components: descriptor, translator, style encoder, and style generator. The descriptor is a multi-head energy-based model that represents a multi-domain image distribution. The components of translator, style encoder, and style generator constitute a diversified image generator. Specifically, given an input image from a source domain, the translator turns it into a stylised output image of the target domain according to a style code, which can be inferred by the style encoder from a reference image or produced by the style generator from a random noise. Since the style generator is represented as an domain-specific distribution of style codes, the translator can provide a one-to-many transformation (i.e., diversified generation) between source domain and target domain. To train our framework, we propose a likelihood-based multi-domain cooperative learning algorithm to jointly train the multi-domain descriptor and the diversified image generator (including translator, style encoder, and style generator modules) via multi-domain MCMC teaching, in which the descriptor guides the diversified image generator to shift its probability density toward the data distribution, while the diversified image generator uses its randomly translated images to initialize the descriptor's Langevin dynamics process for efficient sampling.

Weinan Song, Haoxin Zheng, Dezhan Tu et al.

3D reconstruction of medical imaging from 2D images has become an increasingly interesting topic with the development of deep learning models in recent years. Previous studies in 3D reconstruction from limited X-ray images mainly rely on learning from paired 2D and 3D images, where the reconstruction quality relies on the scale and variation of collected data. This has brought significant challenges in the collection of training data, as only a tiny fraction of patients take two types of radiation examinations in the same period. Although simulation from higher-dimension images could solve this problem, the variance between real and simulated data could bring great uncertainty at the same time. In oral reconstruction, the situation becomes more challenging as only a single panoramic X-ray image is available, where models need to infer the curved shape by prior individual knowledge. To overcome these limitations, we propose Oral-3Dv2 to solve this cross-dimension translation problem in dental healthcare by learning solely on projection information, i.e., the projection image and trajectory of the X-ray tube. Our model learns to represent the 3D oral structure in an implicit way by mapping 2D coordinates into density values of voxels in the 3D space. To improve efficiency and effectiveness, we utilize a multi-head model that predicts a bunch of voxel values in 3D space simultaneously from a 2D coordinate in the axial plane and the dynamic sampling strategy to refine details of the density distribution in the reconstruction result. Extensive experiments in simulated and real data show that our model significantly outperforms existing state-of-the-art models without learning from paired images or prior individual knowledge. To the best of our knowledge, this is the first work of a non-adversarial-learning-based model in 3D radiology reconstruction from a single panoramic X-ray image.

Zehua Chen, Xu Tan, Ke Wang et al.

Denoising diffusion probabilistic models (diffusion models for short) require a large number of iterations in inference to achieve the generation quality that matches or surpasses the state-of-the-art generative models, which invariably results in slow inference speed. Previous approaches aim to optimize the choice of inference schedule over a few iterations to speed up inference. However, this results in reduced generation quality, mainly because the inference process is optimized separately, without jointly optimizing with the training process. In this paper, we propose InferGrad, a diffusion model for vocoder that incorporates inference process into training, to reduce the inference iterations while maintaining high generation quality. More specifically, during training, we generate data from random noise through a reverse process under inference schedules with a few iterations, and impose a loss to minimize the gap between the generated and ground-truth data samples. Then, unlike existing approaches, the training of InferGrad considers the inference process. The advantages of InferGrad are demonstrated through experiments on the LJSpeech dataset showing that InferGrad achieves better voice quality than the baseline WaveGrad under same conditions while maintaining the same voice quality as the baseline but with $3$x speedup ($2$ iterations for InferGrad vs $6$ iterations for WaveGrad).