arXiv subjects

Zhisheng Wang

Publications and source records attributed to Zhisheng Wang.

At least 19 recordsLinked to original sources

arXiv2026

FineCombo-TTS: Collaborative and Precise Controllable Speech Synthesis Using Text Descriptions and Reference Speech

Controllable text-to-speech (TTS) has become a key research focus. However, methods based on either reference speech or text descriptions lack flexibility and precise control, and recent joint approaches remain loosely coupled, with speech modeling timbre and text controlling global style. We propose FineCombo-TTS, a unified framework for speech synthesis grounded in reference speech and guided by text descriptions, enabling flexible and precise control over acoustic attributes. Instead of explicit attribute disentanglement, we learn a unified acoustic representation and introduce a Conditional Flow Matching (CFM)-based Speech Variance Predictor to model fine-grained reference-to-target transformations guided by text descriptions. To support relative attribute control, we construct FineEdit, a structured paired dataset that explicitly encodes source-to-target attribute variations. Experiments demonstrate that our approach achieves flexible, precise, and expressive controllable TTS.

cs.SD↗

arXiv2026

TellWhisper: Tell Whisper Who Speaks When

Multi-speaker automatic speech recognition (MASR) aims to predict ''who spoke when and what'' from multi-speaker speech, a key technology for multi-party dialogue understanding. However, most existing approaches decouple temporal modeling and speaker modeling when addressing ''when'' and ''who'': some inject speaker cues before encoding (e.g., speaker masking), which can cause irreversible information loss; others fuse identity by mixing speaker posteriors after encoding, which may entangle acoustic content with speaker identity. This separation is brittle under rapid turn-taking and overlapping speech, often leading to degraded performance. To address these limitations, we propose TellWhisper, a unified framework that jointly models speaker identity and temporal within the speech encoder. Specifically, we design TS-RoPE, a time-speaker rotary positional encoding: time coordinates are derived from frame indices, while speaker coordinates are derived from speaker activity and pause cues. By applying region-specific rotation angles, the model explicitly captures per-speaker continuity, speaker-turn transitions, and state dynamics, enabling the attention mechanism to simultaneously attend to ''when'' and ''who''. Moreover, to estimate frame-level speaker activity, we develop Hyper-SD, which casts speaker classification in hyperbolic space to enhance inter-class separation and refine speaker-activity estimates. Extensive experiments demonstrate the effectiveness of the proposed approach.

eess.AS↗

arXiv2025

HPSU: A Benchmark for Human-Level Perception in Real-World Spoken Speech Understanding

Recent advances in Speech Large Language Models (Speech LLMs) have led to great progress in speech understanding tasks such as Automatic Speech Recognition (ASR) and Speech Emotion Recognition (SER). However, whether these models can achieve human-level auditory perception, particularly in terms of their ability to comprehend latent intentions and implicit emotions in real-world spoken language, remains underexplored. To this end, we introduce the Human-level Perception in Spoken Speech Understanding (HPSU), a new benchmark for fully evaluating the human-level perceptual and understanding capabilities of Speech LLMs. HPSU comprises over 20,000 expert-validated spoken language understanding samples in English and Chinese. It establishes a comprehensive evaluation framework by encompassing a spectrum of tasks, ranging from basic speaker attribute recognition to complex inference of latent intentions and implicit emotions. To address the issues of data scarcity and high cost of manual annotation in real-world scenarios, we developed a semi-automatic annotation process. This process fuses audio, textual, and visual information to enable precise speech understanding and labeling, thus enhancing both annotation efficiency and quality. We systematically evaluate various open-source and proprietary Speech LLMs. The results demonstrate that even top-performing models still fall considerably short of human capabilities in understanding genuine spoken interactions. Consequently, HPSU will be useful for guiding the development of Speech LLMs toward human-level perception and cognition.

cs.SD↗

arXiv2025

Multi-Reward GRPO for Stable and Prosodic Single-Codebook TTS LLMs at Scale

Recent advances in Large Language Models (LLMs) have transformed text-to-speech (TTS) synthesis, inspiring autoregressive frameworks that represent speech as sequences of discrete codec tokens. Among them, single-codebook TTS LLMs have emerged as compact and streamable architectures that jointly model semantic and acoustic integration. However, despite their efficiency, these models often exhibit unstable prosody, speaker drift, and degraded naturalness. To address these issues, we propose a multi-reward Group Relative Policy Optimization (GRPO) framework that directly optimizes the token generation policy of single-codebook TTS LLMs. Beyond standard intelligibility and speaker similarity objectives, our design integrates three rule-based rewards: a length penalty for duration consistency, an entropy regularization reward for decoding stability, and an LLM-annotated prosody alignment reward that explicitly supervises rhythm. In this prosody reward, an external reasoning LLM predicts multiple plausible pause structures via in-context learning, providing a human-preference-aligned supervisory signal for GRPO training. To assess universality, we further attach a flow-matching (FM) decoder on top of the GRPO-optimized AR backbone and observe consistent additional gains, indicating that our reinforcement optimization enhances the intrinsic AR policy. We further conduct a scalability analysis across data sizes and model scales, revealing that the proposed method consistently enhances prosodic stability, speaker similarity, and overall speech naturalness in single-codebook TTS LLMs.

cs.SD↗

arXiv2024

Optimizing Neural Speech Codec for Low-Bitrate Compression via Multi-Scale Encoding

Neural speech codecs have demonstrated their ability to compress high-quality speech and audio by converting them into discrete token representations. Most existing methods utilize Residual Vector Quantization (RVQ) to encode speech into multiple layers of discrete codes with uniform time scales. However, this strategy overlooks the differences in information density across various speech features, leading to redundant encoding of sparse information, which limits the performance of these methods at low bitrate. This paper proposes MsCodec, a novel multi-scale neural speech codec that encodes speech into multiple layers of discrete codes, each corresponding to a different time scale. This encourages the model to decouple speech features according to their diverse information densities, consequently enhancing the performance of speech compression. Furthermore, we incorporate mutual information loss to augment the diversity among speech codes across different layers. Experimental results indicate that our proposed method significantly improves codec performance at low bitrate.

cs.SD↗

arXiv2024

Geometric Artifact Correction for Symmetric Multi-Linear Trajectory CT: Theory, Method, and Generalization

For extending CT field-of-view to perform non-destructive testing, the Symmetric Multi-Linear trajectory Computed Tomography (SMLCT) has been developed as a successful example of non-standard CT scanning modes. However, inevitable geometric errors can cause severe artifacts in the reconstructed images. The existing calibration method for SMLCT is both crude and inefficient. It involves reconstructing hundreds of images by exhaustively substituting each potential error, and then manually identifying the images with the fewest geometric artifacts to estimate the final geometric errors for calibration. In this paper, we comprehensively and efficiently address the challenging geometric artifacts in SMLCT, , and the corresponding works mainly involve theory, method, and generalization. In particular, after identifying sensitive parameters and conducting some theory analysis of geometric artifacts, we summarize several key properties between sensitive geometric parameters and artifact characteristics. Then, we further construct mathematical relationships that relate sensitive geometric errors to the pixel offsets of reconstruction images with artifact characteristics. To accurately extract pixel bias, we innovatively adapt the Generalized Cross-Correlation with Phase Transform (GCC-PHAT) algorithm, commonly used in sound processing, for our image registration task for each paired symmetric LCT. This adaptation leads to the design of a highly efficient rigid translation registration method. Simulation and physical experiments have validated the excellent performance of this work. Additionally, our results demonstrate significant generalization to common rotated CT and a variant of SMLCT.

cs.CV↗

arXiv2024

Spontaneous Style Text-to-Speech Synthesis with Controllable Spontaneous Behaviors Based on Language Models

Spontaneous style speech synthesis, which aims to generate human-like speech, often encounters challenges due to the scarcity of high-quality data and limitations in model capabilities. Recent language model-based TTS systems can be trained on large, diverse, and low-quality speech datasets, resulting in highly natural synthesized speech. However, they are limited by the difficulty of simulating various spontaneous behaviors and capturing prosody variations in spontaneous speech. In this paper, we propose a novel spontaneous speech synthesis system based on language models. We systematically categorize and uniformly model diverse spontaneous behaviors. Moreover, fine-grained prosody modeling is introduced to enhance the model's ability to capture subtle prosody variations in spontaneous speech.Experimental results show that our proposed method significantly outperforms the baseline methods in terms of prosody naturalness and spontaneous behavior naturalness.

cs.SD↗

arXiv2024

AniPortrait: Audio-Driven Synthesis of Photorealistic Portrait Animation

In this study, we propose AniPortrait, a novel framework for generating high-quality animation driven by audio and a reference portrait image. Our methodology is divided into two stages. Initially, we extract 3D intermediate representations from audio and project them into a sequence of 2D facial landmarks. Subsequently, we employ a robust diffusion model, coupled with a motion module, to convert the landmark sequence into photorealistic and temporally consistent portrait animation. Experimental results demonstrate the superiority of AniPortrait in terms of facial naturalness, pose diversity, and visual quality, thereby offering an enhanced perceptual experience. Moreover, our methodology exhibits considerable potential in terms of flexibility and controllability, which can be effectively applied in areas such as facial motion editing or face reenactment. We release code and model weights at https://github.com/scutzzj/AniPortrait

cs.CV↗

arXiv2024

3D Visibility-aware Generalizable Neural Radiance Fields for Interacting Hands

Neural radiance fields (NeRFs) are promising 3D representations for scenes, objects, and humans. However, most existing methods require multi-view inputs and per-scene training, which limits their real-life applications. Moreover, current methods focus on single-subject cases, leaving scenes of interacting hands that involve severe inter-hand occlusions and challenging view variations remain unsolved. To tackle these issues, this paper proposes a generalizable visibility-aware NeRF (VA-NeRF) framework for interacting hands. Specifically, given an image of interacting hands as input, our VA-NeRF first obtains a mesh-based representation of hands and extracts their corresponding geometric and textural features. Subsequently, a feature fusion module that exploits the visibility of query points and mesh vertices is introduced to adaptively merge features of both hands, enabling the recovery of features in unseen areas. Additionally, our VA-NeRF is optimized together with a novel discriminator within an adversarial learning paradigm. In contrast to conventional discriminators that predict a single real/fake label for the synthesized image, the proposed discriminator generates a pixel-wise visibility map, providing fine-grained supervision for unseen areas and encouraging the VA-NeRF to improve the visual quality of synthesized images. Experiments on the Interhand2.6M dataset demonstrate that our proposed VA-NeRF outperforms conventional NeRFs significantly. Project Page: \url{https://github.com/XuanHuang0/VANeRF}.

cs.CV↗

arXiv2023

Monocular 3D Hand Mesh Recovery via Dual Noise Estimation

Current parametric models have made notable progress in 3D hand pose and shape estimation. However, due to the fixed hand topology and complex hand poses, current models are hard to generate meshes that are aligned with the image well. To tackle this issue, we introduce a dual noise estimation method in this paper. Given a single-view image as input, we first adopt a baseline parametric regressor to obtain the coarse hand meshes. We assume the mesh vertices and their image-plane projections are noisy, and can be associated in a unified probabilistic model. We then learn the distributions of noise to refine mesh vertices and their projections. The refined vertices are further utilized to refine camera parameters in a closed-form manner. Consequently, our method obtains well-aligned and high-quality 3D hand meshes. Extensive experiments on the large-scale Interhand2.6M dataset demonstrate that the proposed method not only improves the performance of its baseline by more than 10$\%$ but also achieves state-of-the-art performance. Project page: \url{https://github.com/hanhuili/DNE4Hand}.

cs.CV↗

arXiv2023

BPF Algorithms for Multiple Source-Translation Computed Tomography Reconstruction

Micro-computed tomography (micro-CT) is a widely used state-of-the-art instrument employed to study the morphological structures of objects in various fields. However, its small field-of-view (FOV) cannot meet the pressing demand for imaging relatively large objects at high spatial resolutions. Recently, we devised a novel scanning mode called multiple source translation CT (mSTCT) that effectively enlarges the FOV of the micro-CT and correspondingly developed a virtual projection-based filtered backprojection (V-FBP) algorithm for reconstruction. Although V-FBP skillfully solves the truncation problem in mSTCT, it requires densely sampled projections to arrive at high-resolution reconstruction, which reduces imaging efficiency. In this paper, we developed two backprojection-filtration (BPF)-based algorithms for mSTCT, i.e., S-BPF (derivatives along source) and D-BPF (derivatives along detector). D-BPF can achieve high-resolution reconstruction with fewer projections than V-FBP and S-BPF. Through simulated and real experiments conducted in this paper, we demonstrate that D-BPF can reduce source sampling by 75% compared with V-FBP at the same spatial resolution, which makes mSTCT more feasible in practice. Meanwhile, S-BPF can yield more stable results than D-BPF, which is similar to V-FBP.

cs.CV↗

arXiv2023

OSNet & MNetO: Two Types of General Reconstruction Architectures for Linear Computed Tomography in Multi-Scenarios

Recently, linear computed tomography (LCT) systems have actively attracted attention. To weaken projection truncation and image the region of interest (ROI) for LCT, the backprojection filtration (BPF) algorithm is an effective solution. However, in BPF for LCT, it is difficult to achieve stable interior reconstruction, and for differentiated backprojection (DBP) images of LCT, multiple rotation-finite inversion of Hilbert transform (Hilbert filtering)-inverse rotation operations will blur the image. To satisfy multiple reconstruction scenarios for LCT, including interior ROI, complete object, and exterior region beyond field-of-view (FOV), and avoid the rotation operations of Hilbert filtering, we propose two types of reconstruction architectures. The first overlays multiple DBP images to obtain a complete DBP image, then uses a network to learn the overlying Hilbert filtering function, referred to as the Overlay-Single Network (OSNet). The second uses multiple networks to train different directional Hilbert filtering models for DBP images of multiple linear scannings, respectively, and then overlays the reconstructed results, i.e., Multiple Networks Overlaying (MNetO). In two architectures, we introduce a Swin Transformer (ST) block to the generator of pix2pixGAN to extract both local and global features from DBP images at the same time. We investigate two architectures from different networks, FOV sizes, pixel sizes, number of projections, geometric magnification, and processing time. Experimental results show that two architectures can both recover images. OSNet outperforms BPF in various scenarios. For the different networks, ST-pix2pixGAN is superior to pix2pixGAN and CycleGAN. MNetO exhibits a few artifacts due to the differences among the multiple models, but any one of its models is suitable for imaging the exterior edge in a certain direction.

cs.CV↗

arXiv2023

ExpCLIP: Bridging Text and Facial Expressions via Semantic Alignment

The objective of stylized speech-driven facial animation is to create animations that encapsulate specific emotional expressions. Existing methods often depend on pre-established emotional labels or facial expression templates, which may limit the necessary flexibility for accurately conveying user intent. In this research, we introduce a technique that enables the control of arbitrary styles by leveraging natural language as emotion prompts. This technique presents benefits in terms of both flexibility and user-friendliness. To realize this objective, we initially construct a Text-Expression Alignment Dataset (TEAD), wherein each facial expression is paired with several prompt-like descriptions.We propose an innovative automatic annotation method, supported by Large Language Models (LLMs), to expedite the dataset construction, thereby eliminating the substantial expense of manual annotation. Following this, we utilize TEAD to train a CLIP-based model, termed ExpCLIP, which encodes text and facial expressions into semantically aligned style embeddings. The embeddings are subsequently integrated into the facial animation generator to yield expressive and controllable facial animations. Given the limited diversity of facial emotions in existing speech-driven facial animation training data, we further introduce an effective Expression Prompt Augmentation (EPA) mechanism to enable the animation generator to support unprecedented richness in style control. Comprehensive experiments illustrate that our method accomplishes expressive facial animation generation and offers enhanced flexibility in effectively conveying the desired style.

cs.CV↗

arXiv2023

LongDanceDiff: Long-term Dance Generation with Conditional Diffusion Model

Dancing with music is always an essential human art form to express emotion. Due to the high temporal-spacial complexity, long-term 3D realist dance generation synchronized with music is challenging. Existing methods suffer from the freezing problem when generating long-term dances due to error accumulation and training-inference discrepancy. To address this, we design a conditional diffusion model, LongDanceDiff, for this sequence-to-sequence long-term dance generation, addressing the challenges of temporal coherency and spatial constraint. LongDanceDiff contains a transformer-based diffusion model, where the input is a concatenation of music, past motions, and noised future motions. This partial noising strategy leverages the full-attention mechanism and learns the dependencies among music and past motions. To enhance the diversity of generated dance motions and mitigate the freezing problem, we introduce a mutual information minimization objective that regularizes the dependency between past and future motions. We also address common visual quality issues in dance generation, such as foot sliding and unsmooth motion, by incorporating spatial constraints through a Global-Trajectory Modulation (GTM) layer and motion perceptual losses, thereby improving the smoothness and naturalness of motion generation. Extensive experiments demonstrate a significant improvement in our approach over the existing state-of-the-art methods. We plan to release our codes and models soon.

cs.CV↗

arXiv2023

Analytical reconstructions of full-scan multiple source-translation computed tomography under large field of views

This paper is to investigate the high-quality analytical reconstructions of multiple source-translation computed tomography (mSTCT) under an extended field of view (FOV). Under the larger FOVs, the previously proposed backprojection filtration (BPF) algorithms for mSTCT, including D-BPF and S-BPF (their differences are different derivate directions along the detector and source, respectively), make some errors and artifacts in the reconstructed images due to a backprojection weighting factor and the half-scan mode, which deviates from the intention of mSTCT imaging. In this paper, to achieve reconstruction with as little error as possible under the extremely extended FOV, we combine the full-scan mSTCT (F-mSTCT) geometry with the previous BPF algorithms to study the performance and derive a suitable redundancy-weighted function for F-mSTCT. The experimental results indicate FS-BPF can get high-quality, stable images under the extremely extended FOV of imaging a large object, though it requires more projections than FD-BPF. Finally, for different practical requirements in extending FOV imaging, we give suggestions on algorithm selection.

cs.CV↗

arXiv2022

AccentSpeech: Learning Accent from Crowd-sourced Data for Target Speaker TTS with Accents

Learning accent from crowd-sourced data is a feasible way to achieve a target speaker TTS system that can synthesize accent speech. To this end, there are two challenging problems to be solved. First, direct use of the poor acoustic quality crowd-sourced data and the target speaker data in accent transfer will apparently lead to synthetic speech with degraded quality. To mitigate this problem, we take a bottleneck feature (BN) based TTS approach, in which TTS is decomposed into a Text-to-BN (T2BN) module to learn accent and a BN-to-Mel (BN2Mel) module to learn speaker timbre, where neural network based BN feature serves as the intermediate representation that are robust to noise interference. Second, direct training T2BN using the crowd-sourced data in the two-stage system will produce accent speech of target speaker with poor prosody. This is because the the crowd-sourced recordings are contributed from the ordinary unprofessional speakers. To tackle this problem, we update the two-stage approach to a novel three-stage approach, where T2BN and BN2Mel are trained using the high-quality target speaker data and a new BN-to-BN module is plugged in between the two modules to perform accent transfer. To train the BN2BN module, the parallel unaccented and accented BN features are obtained by a proposed data augmentation procedure. Finally the proposed three-stage approach manages to produce accent speech for the target speaker with good prosody, as the prosody pattern is inherited from the professional target speaker and accent transfer is achieved by the BN2BN module at the same time. The proposed approach, named as AccentSpeech, is validated in a Mandarin TTS accent transfer task.

cs.SD↗

arXiv2021

Cross-lingual Text Classification with Heterogeneous Graph Neural Network

Cross-lingual text classification aims at training a classifier on the source language and transferring the knowledge to target languages, which is very useful for low-resource languages. Recent multilingual pretrained language models (mPLM) achieve impressive results in cross-lingual classification tasks, but rarely consider factors beyond semantic similarity, causing performance degradation between some language pairs. In this paper we propose a simple yet effective method to incorporate heterogeneous information within and across languages for cross-lingual text classification using graph convolutional networks (GCN). In particular, we construct a heterogeneous graph by treating documents and words as nodes, and linking nodes with different relations, which include part-of-speech roles, semantic similarity, and document translations. Extensive experiments show that our graph-based method significantly outperforms state-of-the-art models on all tasks, and also achieves consistent performance gain over baselines in low-resource settings where external tools like translators are unavailable.

cs.CL↗

arXiv2019

Training Deep Neural Networks Using Posit Number System

With the increasing size of Deep Neural Network (DNN) models, the high memory space requirements and computational complexity have become an obstacle for efficient DNN implementations. To ease this problem, using reduced-precision representations for DNN training and inference has attracted many interests from researchers. This paper first proposes a methodology for training DNNs with the posit arithmetic, a type- 3 universal number (Unum) format that is similar to the floating point(FP) but has reduced precision. A warm-up training strategy and layer-wise scaling factors are adopted to stabilize training and fit the dynamic range of DNN parameters. With the proposed training methodology, we demonstrate the first successful training of DNN models on ImageNet image classification task in 16 bits posit with no accuracy loss. Then, an efficient hardware architecture for the posit multiply-and-accumulate operation is also proposed, which can achieve significant improvement in energy efficiency than traditional floating-point implementations. The proposed design is helpful for future low-power DNN training accelerators.

cs.LG↗