arXiv subjects
Heng Xie
Publications and source records attributed to Heng Xie.
M3-TTS: Multi-modal DiT Alignment & Mel-latent for Zero-shot High-fidelity Speech Synthesis
Non-autoregressive (NAR) text-to-speech synthesis relies on length alignment between text sequences and audio representations, constraining naturalness and expressiveness. Existing methods depend on duration modeling or pseudo-alignment strategies that severely limit naturalness and computational efficiency. We propose M3-TTS, a concise and efficient NAR TTS paradigm based on multi-modal diffusion transformer (MM-DiT) architecture. M3-TTS employs joint diffusion transformer layers for cross-modal alignment, achieving stable monotonic alignment between variable-length text-speech sequences without pseudo-alignment requirements. Single diffusion transformer layers further enhance acoustic detail modeling. The framework integrates a mel-vae codec that provides 3* training acceleration. Experimental results on Seed-TTS and AISHELL-3 benchmarks demonstrate that M3-TTS achieves state-of-the-art NAR performance with the lowest word error rates (1.36\% English, 1.31\% Chinese) while maintaining competitive naturalness scores. Code and demos will be available at https://wwwwxp.github.io/M3-TTS.
PSA-MF: Personality-Sentiment Aligned Multi-Level Fusion for Multimodal Sentiment Analysis
Multimodal sentiment analysis (MSA) is a research field that recognizes human sentiments by combining textual, visual, and audio modalities. The main challenge lies in integrating sentiment-related information from different modalities, which typically arises during the unimodal feature extraction phase and the multimodal feature fusion phase. Existing methods extract only shallow information from unimodal features during the extraction phase, neglecting sentimental differences across different personalities. During the fusion phase, they directly merge the feature information from each modality without considering differences at the feature level. This ultimately affects the model's recognition performance. To address this problem, we propose a personality-sentiment aligned multi-level fusion framework. We introduce personality traits during the feature extraction phase and propose a novel personality-sentiment alignment method to obtain personalized sentiment embeddings from the textual modality for the first time. In the fusion phase, we introduce a novel multi-level fusion method. This method gradually integrates sentimental information from textual, visual, and audio modalities through multimodal pre-fusion and a multi-level enhanced fusion strategy. Our method has been evaluated through multiple experiments on two commonly used datasets, achieving state-of-the-art results.
Hermitian K-theory of Lagrangian Grassmannians via reducible Gorenstein models
We construct a family of moduli spaces, called generalized Lagrangian flag schemes, that are reducible Gorenstein (hence singular), and that admit well-behaved pushforward and pullback operations in Hermitian $K$-theory. These schemes arise naturally in our computations. Using them, we prove that the Hermitian $K$-theory of a Lagrangian Grassmannian over a regular base splits as a direct sum of copies of the base's (Hermitian) $K$-theory, indexed by certain shifted Young diagrams. The isomorphism is realized via pullback to each generalized Lagrangian flag scheme followed by pushforward to the Lagrangian Grassmannian. This yields an unusual example in which both the base and the target are regular schemes, while the intermediate reducible Gorenstein models remain sufficient to allow explicit computations in Hermitian K-theory of regular schemes.
MDPE: A Multimodal Deception Dataset with Personality and Emotional Characteristics
Deception detection has garnered increasing attention in recent years due to the significant growth of digital media and heightened ethical and security concerns. It has been extensively studied using multimodal methods, including video, audio, and text. In addition, individual differences in deception production and detection are believed to play a crucial role.Although some studies have utilized individual information such as personality traits to enhance the performance of deception detection, current systems remain limited, partly due to a lack of sufficient datasets for evaluating performance. To address this issue, we introduce a multimodal deception dataset MDPE. Besides deception features, this dataset also includes individual differences information in personality and emotional expression characteristics. It can explore the impact of individual differences on deception behavior. It comprises over 104 hours of deception and emotional videos from 193 subjects. Furthermore, we conducted numerous experiments to provide valuable insights for future deception detection research. MDPE not only supports deception detection, but also provides conditions for tasks such as personality recognition and emotion recognition, and can even study the relationships between them. We believe that MDPE will become a valuable resource for promoting research in the field of affective computing.
The connecting homomorphism for Hermitian $K$-theory
We provide a geometric interpretation for the connecting homomorphism in the localization sequence of Hermitian $K$-theory. As an application, we compute the Hermitian $K$-theory of projective bundles and Grassmannians in the regular case. We provide an explicit basis for Hermitian $K$-theory of Grassmannians, which is indexed by even Young diagrams together with another special class of Young diagrams, that we call $\textit{buffalo-check}$ Young diagrams. To achieve this, we develop pushforwards and pullbacks in Hermitian $K$-theory using Grothendieck's residue complexes, and we establish fundamental theorems for those pushforwards and pullbacks, including base change, projection, and excess intersection formulas.
Witt groups of spinor varieties
We show that Witt groups of spinor varieties (aka.\ maximal isotropic Grassmannians) can be presented by combinatorial objects called even shifted young diagram. Our method relies on the Blow-up setup of Balmer-Calmès, and we investigate the connecting homomorphism of the localization sequence via the projective bundle formula of Walter-Nenashev, the projection formula of Calmès-Hornbostel and the excess intersection formula of Fasel.
The real cycle class map
The classical cycle class map for a smooth complex variety sends cycles in the Chow ring to cycles in the singular cohomology ring. We study two cycle class maps for smooth real varieties: the map from the I-cohomology ring to singular cohomology induced by the signature, and a new cycle class map defined on the Chow-Witt ring. For both maps, we establish basic compatibility results like compatibility with pullbacks, pushforwards and cup products. As a first application of these general results, we show that both cycle class maps are isomorphisms for cellular varieties.
Towards Cooperative Transport of a Suspended Payload via Two Aerial Robots with Inertial Sensing
This paper addresses the problem of cooperative transport of a point mass hoisted by two aerial robots. Treating the robots as a leader and a follower, the follower stabilizes the system with respect to the leader using only feedback from its Inertial Measurement Units (IMU). This is accomplished by neglecting the acceleration of the leader, analyzing the system through the generalized coordinates or the cables' angles, and employing an observation model based on the IMU measurements. A lightweight estimator based on an Extended Kalman Filter (EKF) and a controller are derived to stabilize the robot-payload-robot system. The proposed methods are verified with extensive flight experiments, first with a single robot and then with two robots. The results show that the follower is capable of realizing the desired quasi-static trajectory using only its IMU measurements. The outcomes demonstrate promising progress towards the goal of autonomous cooperative transport of a suspended payload via small flying robots with minimal sensing and computational requirements.
A Gersten complex on real schemes
We discuss a connection between coherent duality and Verdier duality via a Gersten-type complex of sheaves on real schemes, and show that this construction gives a dualizing object in the derived category, which is compatible with the exceptional inverse image functor $f^!$. The hypercohomology of this complex coincides with hypercohomology of the sheafified Gersten-Witt complex, which in some cases can be related to topological or semialgebraic Borel-Moore homology.
A Transfer morphism for Hermitian K-theory of schemes with involution
In this paper, we consider the Hermitian $K$-theory of schemes with involution, for which we construct a transfer morphism and prove a version of the dévissage theorem. This theorem is then used to compute the Hermitian $K$-theory of $\mathbb{P}^1$ with involution given by $[X:Y] \mapsto [Y:X]$. We also prove the $C_2$-equivariant $\A^1$-invariance of Hermitian $K$-theory, which confirms the representability of Hermitian $K$-theory in the $C_2$-equivariant motivic homotopy category of Heller, Krishna and \{O}stvær \cite{HKO14}.
Chow-Witt rings of split quadrics
We compute the Chow-Witt rings of split quadrics over a field of characteristic not two. We even determine the full bigraded I-cohomology and Milnor-Witt cohomology rings, including twists by line bundles. The results on I-cohomology corroborate the general philosophy that I-cohomology is an algebro-geometric version of singular cohomology of real varieties: our explicit calculations confirm that the I-cohomology ring of a split quadric over the reals is isomorphic to the singular cohomology ring of the space of its real points.
Hermitian K-theory of quadric hypersurfaces
Let k be a commutative ring in which 2 is invertible. We prove that the Hermitian K-theory of quadric hypersurfaces over k admits fibration sequences relating it to the base ring and to Clifford algebras equipped with various duality coefficients. All shifts and twists are taken into account.