Search

Recent advancements in large language models (LLMs) and multimodal speech-text models have laid the groundwork for seamless voice interactions, enabling real-time, natural, and human-like conversations. Previous models for voice interactions are categorized as native and aligned. Native models integrate speech and text processing in one framework but struggle with issues like differing sequence lengths and insufficient pre-training. Aligned models maintain text LLM capabilities but are often limited by small datasets and a narrow focus on speech tasks. In this work, we introduce MinMo, a Multimodal Large Language Model with approximately 8B parameters for seamless voice interaction. We address the main limitations of prior aligned multimodal models. We train MinMo through multiple stages of speech-to-text alignment, text-to-speech alignment, speech-to-speech alignment, and duplex interaction alignment, on 1.4 million hours of diverse speech data and a broad range of speech tasks. After the multi-stage training, MinMo achieves state-of-the-art performance across various benchmarks for voice comprehension and generation while maintaining the capabilities of text LLMs, and also facilitates full-duplex conversation, that is, simultaneous two-way communication between the user and the system. Moreover, we propose a novel and simple voice decoder that outperforms prior models in voice generation. The enhanced instruction-following capabilities of MinMo supports controlling speech generation based on user instructions, with various nuances including emotions, dialects, and speaking rates, and mimicking specific voices. For MinMo, the speech-to-text latency is approximately 100ms, full-duplex latency is approximately 600ms in theory and 800ms in practice. The MinMo project web page is https://funaudiollm.github.io/minmo, and the code and models will be released soon., Comment: Work in progress. Authors are listed in alphabetical order by family name

In sound event detection (SED), overlapping sound events pose a significant challenge, as certain events can be easily masked by background noise or other events, resulting in poor detection performance. To address this issue, we propose the text-queried SED (TQ-SED) framework. Specifically, we first pre-train a language-queried audio source separation (LASS) model to separate the audio tracks corresponding to different events from the input audio. Then, multiple target SED branches are employed to detect individual events. AudioSep is a state-of-the-art LASS model, but has limitations in extracting dynamic audio information because of its pure convolutional structure for separation. To address this, we integrate a dual-path recurrent neural network block into the model. We refer to this structure as AudioSep-DP, which achieves the first place in DCASE 2024 Task 9 on language-queried audio source separation (objective single model track). Experimental results show that TQ-SED can significantly improve the SED performance, with an improvement of 7.22\% on F1 score over the conventional framework. Additionally, we setup comprehensive experiments to explore the impact of model complexity. The source code and pre-trained model are released at https://github.com/apple-yinhan/TQ-SED., Comment: Accepted by ICASSP 2025

Speaker diarization, the process of segmenting an audio stream or transcribed speech content into homogenous partitions based on speaker identity, plays a crucial role in the interpretation and analysis of human speech. Most existing speaker diarization systems rely exclusively on unimodal acoustic information, making the task particularly challenging due to the innate ambiguities of audio signals. Recent studies have made tremendous efforts towards audio-visual or audio-semantic modeling to enhance performance. However, even the incorporation of up to two modalities often falls short in addressing the complexities of spontaneous and unstructured conversations. To exploit more meaningful dialogue patterns, we propose a novel multimodal approach that jointly utilizes audio, visual, and semantic cues to enhance speaker diarization. Our method elegantly formulates the multimodal modeling as a constrained optimization problem. First, we build insights into the visual connections among active speakers and the semantic interactions within spoken content, thereby establishing abundant pairwise constraints. Then we introduce a joint pairwise constraint propagation algorithm to cluster speakers based on these visual and semantic constraints. This integration effectively leverages the complementary strengths of different modalities, refining the affinity estimation between individual speaker embeddings. Extensive experiments conducted on multiple multimodal datasets demonstrate that our approach consistently outperforms state-of-the-art speaker diarization methods.

Training speaker-discriminative and robust speaker verification systems without explicit speaker labels remains a persisting challenge. In this paper, we propose a new self-supervised speaker verification approach, Self-Distillation Prototypes Network (SDPN), which effectively facilitates self-supervised speaker representation learning. SDPN assigns the representation of the augmented views of an utterance to the same prototypes as the representation of the original view, thereby enabling effective knowledge transfer between the views. Originally, due to the lack of negative pairs in the SDPN training process, the network tends to align positive pairs very closely in the embedding space, a phenomenon known as model collapse. To alleviate this problem, we introduce a diversity regularization term to embeddings in SDPN. Comprehensive experiments on the VoxCeleb datasets demonstrate the superiority of SDPN in self-supervised speaker verification. SDPN sets a new state-of-the-art on the VoxCeleb1 speaker verification evaluation benchmark, achieving Equal Error Rate 1.80%, 1.99%, and 3.62% for trial VoxCeleb1-O, VoxCeleb1-E and VoxCeleb1-H respectively, without using any speaker labels in training., Comment: We update this paper to an earlier paper arXiv:2308.02774

Speaker verification systems experience significant performance degradation when tasked with short-duration trial recordings. To address this challenge, a multi-scale feature fusion approach has been proposed to effectively capture speaker characteristics from short utterances. Constrained by the model's size, a robust backbone Enhanced Res2Net (ERes2Net) combining global and local feature fusion demonstrates sub-optimal performance in short-duration speaker verification. To further improve the short-duration feature extraction capability of ERes2Net, we expand the channel dimension within each stage. However, this modification also increases the number of model parameters and computational complexity. To alleviate this problem, we propose an improved ERes2NetV2 by pruning redundant structures, ultimately reducing both the model parameters and its computational cost. A range of experiments conducted on the VoxCeleb datasets exhibits the superiority of ERes2NetV2, which achieves EER of 0.61% for the full-duration trial, 0.98% for the 3s-duration trial, and 1.48% for the 2s-duration trial on VoxCeleb1-O, respectively.

We introduce 3D-Speaker-Toolkit, an open-source toolkit for multimodal speaker verification and diarization, designed for meeting the needs of academic researchers and industrial practitioners. The 3D-Speaker-Toolkit adeptly leverages the combined strengths of acoustic, semantic, and visual data, seamlessly fusing these modalities to offer robust speaker recognition capabilities. The acoustic module extracts speaker embeddings from acoustic features, employing both fully-supervised and self-supervised learning approaches. The semantic module leverages advanced language models to comprehend the substance and context of spoken language, thereby augmenting the system's proficiency in distinguishing speakers through linguistic patterns. The visual module applies image processing technologies to scrutinize facial features, which bolsters the precision of speaker diarization in multi-speaker environments. Collectively, these modules empower the 3D-Speaker-Toolkit to achieve substantially improved accuracy and reliability in speaker-related tasks. With 3D-Speaker-Toolkit, we establish a new benchmark for multimodal speaker analysis. The toolkit also includes a handful of open-source state-of-the-art models and a large-scale dataset containing over 10,000 speakers. The toolkit is publicly available at https://github.com/modelscope/3D-Speaker.

Speaker diarization has gained considerable attention within speech processing research community. Mainstream speaker diarization rely primarily on speakers' voice characteristics extracted from acoustic signals and often overlook the potential of semantic information. Considering the fact that speech signals can efficiently convey the content of a speech, it is of our interest to fully exploit these semantic cues utilizing language models. In this work we propose a novel approach to effectively leverage semantic information in clustering-based speaker diarization systems. Firstly, we introduce spoken language understanding modules to extract speaker-related semantic information and utilize these information to construct pairwise constraints. Secondly, we present a novel framework to integrate these constraints into the speaker diarization pipeline, enhancing the performance of the entire system. Extensive experiments conducted on the public dataset demonstrate the consistent superiority of our proposed approach over acoustic-only speaker diarization systems.

Training speaker-discriminative and robust speaker verification systems without explicit speaker labels remains a persistent challenge. In this paper, we propose a novel self-supervised speaker verification approach, Self-Distillation Prototypes Network (SDPN), which effectively facilitates self-supervised speaker representation learning. SDPN assigns the representation of the augmented views of an utterance to the same prototypes as the representation of the original view, thereby enabling effective knowledge transfer between the augmented and original views. Due to lack of negative pairs in the SDPN training process, the network tends to align positive pairs quite closely in the embedding space, a phenomenon known as model collapse. To mitigate this problem, we introduce a diversity regularization term to embeddings in SDPN. Comprehensive experiments on the VoxCeleb datasets demonstrate the superiority of SDPN among self-supervised speaker verification approaches. SDPN sets a new state-of-the-art on the VoxCeleb1 speaker verification evaluation benchmark, achieving Equal Error Rate 1.80%, 1.99%, and 3.62% for trial VoxCeleb1-O, VoxCeleb1-E and VoxCeleb1-H, without using any speaker labels in training. Ablation studies show that both proposed learnable prototypes in self-distillation network and diversity regularization contribute to the verification performance., Comment: arXiv admin note: text overlap with arXiv:2211.04168

Disentangling uncorrelated information in speech utterances is a crucial research topic within speech community. Different speech-related tasks focus on extracting distinct speech representations while minimizing the affects of other uncorrelated information. We present a large-scale speech corpus to facilitate the research of speech representation disentanglement. 3D-Speaker contains over 10,000 speakers, each of whom are simultaneously recorded by multiple Devices, locating at different Distances, and some speakers are speaking multiple Dialects. The controlled combinations of multi-dimensional audio data yield a matrix of a diverse blend of speech representation entanglement, thereby motivating intriguing methods to untangle them. The multi-domain nature of 3D-Speaker also makes it a suitable resource to evaluate large universal speech models and experiment methods of out-of-domain learning and self-supervised learning. https://3dspeaker.github.io/

Speaker diarization(SD) is a classic task in speech processing and is crucial in multi-party scenarios such as meetings and conversations. Current mainstream speaker diarization approaches consider acoustic information only, which result in performance degradation when encountering adverse acoustic conditions. In this paper, we propose methods to extract speaker-related information from semantic content in multi-party meetings, which, as we will show, can further benefit speaker diarization. We introduce two sub-tasks, Dialogue Detection and Speaker-Turn Detection, in which we effectively extract speaker information from conversational semantics. We also propose a simple yet effective algorithm to jointly model acoustic and semantic information and obtain speaker-identified texts. Experiments on both AISHELL-4 and AliMeeting datasets show that our method achieves consistent improvements over acoustic-only speaker diarization systems., Comment: Accepted to Findings of ACL 2023

Effective fusion of multi-scale features is crucial for improving speaker verification performance. While most existing methods aggregate multi-scale features in a layer-wise manner via simple operations, such as summation or concatenation. This paper proposes a novel architecture called Enhanced Res2Net (ERes2Net), which incorporates both local and global feature fusion techniques to improve the performance. The local feature fusion (LFF) fuses the features within one single residual block to extract the local signal. The global feature fusion (GFF) takes acoustic features of different scales as input to aggregate global signal. To facilitate effective feature fusion in both LFF and GFF, an attentional feature fusion module is employed in the ERes2Net architecture, replacing summation or concatenation operations. A range of experiments conducted on the VoxCeleb datasets demonstrate the superiority of the ERes2Net in speaker verification. Code has been made publicly available at https://github.com/alibaba-damo-academy/3D-Speaker.

Time delay neural network (TDNN) has been proven to be efficient for speaker verification. One of its successful variants, ECAPA-TDNN, achieved state-of-the-art performance at the cost of much higher computational complexity and slower inference speed. This makes it inadequate for scenarios with demanding inference rate and limited computational resources. We are thus interested in finding an architecture that can achieve the performance of ECAPA-TDNN and the efficiency of vanilla TDNN. In this paper, we propose an efficient network based on context-aware masking, namely CAM++, which uses densely connected time delay neural network (D-TDNN) as backbone and adopts a novel multi-granularity pooling to capture contextual information at different levels. Extensive experiments on two public benchmarks, VoxCeleb and CN-Celeb, demonstrate that the proposed architecture outperforms other mainstream speaker verification systems with lower computational cost and faster inference speed.

Training robust speaker verification systems without speaker labels has long been a challenging task. Previous studies observed a large performance gap between self-supervised and fully supervised methods. In this paper, we apply a non-contrastive self-supervised learning framework called DIstillation with NO labels (DINO) and propose two regularization terms applied to embeddings in DINO. One regularization term guarantees the diversity of the embeddings, while the other regularization term decorrelates the variables of each embedding. The effectiveness of various data augmentation techniques are explored, on both time and frequency domain. A range of experiments conducted on the VoxCeleb datasets demonstrate the superiority of the regularized DINO framework in speaker verification. Our method achieves the state-of-the-art speaker verification performance under a single-stage self-supervised setting on VoxCeleb. Code has been made publicly available at https://github.com/alibaba-damo-academy/3D-Speaker.

When imaging moving objects, single-pixel imaging produces motion blur. This paper proposes a new single-pixel imaging method, which can achieve anti-motion blur imaging of a fast-moving object. The geometric moment patterns and Hadamard patterns are used to alternately encode the position information and the image information of the object with time-division multiplexing. In the reconstruction process, the object position information is extracted independently and combining motion-compensation reconstruction algorithm to decouple the object motion from image information. As a result, the anti-motion blur image and the high frame rate object positions are obtained. Experimental results show that for a moving object with an angular velocity of up to 0.5rad/s relative to the imaging system, the proposed method achieves a localization frequency of 5.55kHz, and gradually reconstructs a clear image of the fast-moving object with a pseudo resolution of 512x512. The method has application prospects in single-pixel imaging of the fast-moving object., Comment: 13 pages,7figures

Unsupervised clustering on speakers is becoming increasingly important for its potential uses in semi-supervised learning. In reality, we are often presented with enormous amounts of unlabeled data from multi-party meetings and discussions. An effective unsupervised clustering approach would allow us to significantly increase the amount of training data without additional costs for annotations. Recently, methods based on graph convolutional networks (GCN) have received growing attention for unsupervised clustering, as these methods exploit the connectivity patterns between nodes to improve learning performance. In this work, we present a GCN-based approach for semi-supervised learning. Given a pre-trained embedding extractor, a graph convolutional network is trained on the labeled data and clusters unlabeled data with "pseudo-labels". We present a self-correcting training mechanism that iteratively runs the cluster-train-correct process on pseudo-labels. We show that this proposed approach effectively uses unlabeled data and improves speaker recognition accuracy., Comment: Accepted by ICASSP 2022

Second-order photonic topological insulators that host highly localized corner states resilient to defects, are opening new routes towards developing fascinating photonic devices. However, the existing works on second-order photonic topological insulators have mainly focused on either transverse magnetic or transverse electric modes. In this paper, we propose a dual-polarization topological photonic crystal structure for both transverse magnetic and transverse electric modes through topology optimization. Simple tight-binding lattice models are constructed to reveal the topological features of the optimized photonic crystal structure in a transparent way. The optimized dual-polarization second-order photonic topological insulator hosts four groups of corner states with different profiles and eigenfrequencies for both the transverse magnetic and transverse electric modes. Moreover, the robustness of theses corner states against defects is explicitly demonstrated. Our results offer opportunities for developing polarization-independent topological photonic devices., Comment: 9 pages, 9 figures

Topology-driven nonlinear light-matter effects open up new paradigms for both topological photonics and nonlinear optics. Here, we propose to achieve high-efficiency second-harmonic generation in a second-order photonic topological insulator. Such system hosts highly localized topological corner states with large quality factors for both fundamental and second harmonic waves, which could be matched perfectly in frequency by simply tuning the structural parameters. Through the nonlinear interaction of the doubly resonant topological corner states, unprecedented frequency conversion efficiency is predicted. In addition, the robustness of the nonlinear process against defects is also demonstrated. Our work opens up new avenues toward topologically protected nonlinear frequency conversion using high-order photonic topological modes., Comment: 7 pages, 8 figures

Meta-learning has recently become a research hotspot in speaker verification (SV). We introduce two methods to improve the meta-learning training for SV in this paper. For the first method, a backbone embedding network is first jointly trained with the conventional cross entropy loss and prototypical networks (PN) loss. Then, inspired by speaker adaptive training in speech recognition, additional transformation coefficients are trained with only the PN loss. The transformation coefficients are used to modify the original backbone embedding network in the x-vector extraction process. Furthermore, the random erasing data augmentation technique is applied to all support samples in each episode to construct positive pairs, and a contrastive loss between the augmented and the original support samples is added to the objective in model training. Experiments are carried out on the SITW and VOiCES databases. Both of the methods can obtain consistent improvements over existing meta-learning training frameworks. By combining these two methods, we can observe further improvements on these two databases.

In this work, we introduce metric learning (ML) to enhance the deep embedding learning for text-independent speaker verification (SV). Specifically, the deep speaker embedding network is trained with conventional cross entropy loss and auxiliary pair-based ML loss function. For the auxiliary ML task, training samples of a mini-batch are first arranged into pairs, then positive and negative pairs are selected and weighted through their own and relative similarities, and finally the auxiliary ML loss is calculated by the similarity of the selected pairs. To evaluate the proposed method, we conduct experiments on the Speaker in the Wild (SITW) dataset. The results demonstrate the effectiveness of the proposed method., Comment: Not a particularly high-quality work, so we request withdrawal

The universal speech attributes for x-vector based speaker verification (SV) are addressed in this paper. The manner and place of articulation form the fundamental speech attribute unit (SAU), and then new speech attribute (NSA) units for acoustic modeling are generated by tied tri-SAU states. An improved cross-stitch network is adopted as a multitask learning (MTL) framework for integrating these universal speech attributes into the x-vector network training process. Experiments are conducted on common condition 5 (CC5) of the core-core and the 10 s-10 s tests of the NIST SRE10 evaluation set, and the proposed algorithm can achieve consistent improvements over the baseline x-vector on both these tasks., Comment: Not a particularly high-quality work, so we request withdrawal

Topological photonics revolutionizes some of the traditional approaches to light propagation and manipulation, and it provides unprecedented means for developing novel photonic devices. Recently discovered higher-order topological phases go beyond the conventional bulk-edge correspondence for photonic crystals and introduce novel opportunities for topological protection. Here, we introduce an intelligent numerical approach for inverse design of higher-order photonic topological insulators with great flexibility for controlling both topological edge and topological corner states. In particular, we consider the second-order photonic topological insulator and design several structures supporting both edge and corner states at different frequencies. By carefully programming these structures, we suggest a novel approach for topological routing of the edge and corner states by changing the operational frequency. Our finding paves the way to integrated topological photonic devices with novel functionalities., Comment: 12 pages, 5 figures

High-order topological insulators (TIs) develop the conventional bulk-boundary correspondence theory and rise the interest in searching innovative topological materials. To realize a high-order TI with a wide passband of 1D and 2D transportation modes, we design non-trivial and trivial 3D sonic crystals whose combination mimics the Su-Schrieffer-Heeger model. The high-order topological boundary states can be found at the interfaces, including 0D corner state, 1D hinge state, and 2D surface state. The fabricated sample with the bent two-dimensional and one-dimensional acoustic channels exhibits the multidimensional sound propagation in space, and also verifies the transition between the complete band gap, hinge states, and surface states within the bulk band gap. Among them, the bandwidth of the single-mode hinge state achieves a large relative bandwidth 9.1%, in which sound transports one-dimensionally without significant leak into the surfaces or the bulk. The high-order topological states in the study pave the way for multidimensional sound manipulation in space., Comment: 21 pages, 7 figures