Full-Duplex Speech Dialogue Systems (Full-Duplex SDS) have significantly enhanced the naturalness of human-machine interaction by enabling real-time bidirectional communication. However, existing approaches face challenges such as difficulties in independent module optimization and contextual noise interference due to highly coupled architectural designs and oversimplified binary state modeling. This paper proposes FlexDuo, a flexible full-duplex control module that decouples duplex control from spoken dialogue systems through a plug-and-play architectural design. Furthermore, inspired by human information-filtering mechanisms in conversations, we introduce an explicit Idle state. On one hand, the Idle state filters redundant noise and irrelevant audio to enhance dialogue quality. On the other hand, it establishes a semantic integrity-based buffering mechanism, reducing the risk of mutual interruptions while ensuring accurate response transitions. Experimental results on the Fisher corpus demonstrate that FlexDuo reduces the false interruption rate by 24.9% and improves response accuracy by 7.6% compared to integrated full-duplex dialogue system baselines. It also outperforms voice activity detection (VAD) controlled baseline systems in both Chinese and English dialogue quality. The proposed modular architecture and state-based dialogue model provide a novel technical pathway for building flexible and efficient duplex dialogue systems.
Utilizing complex-valued neural networks (CVNNs) in wireless communication tasks has received growing attention for their ability to provide natural and effective representation of complex-valued signals and data. However, existing studies typically employ complex-valued versions of simple neural network architectures. Not only they merely scratch the surface of the extensive range of modern deep learning techniques, theoretical understanding of the superior performance of CVNNs is missing. To this end, this paper aims to fill both the theoretical and practice gap of employing CVNNs in wireless communications. In particular, we provide a comprehensive description on the various operations in CVNNs and theoretically prove that the CVNN requires fewer layers than the real-valued counterpart to achieve a given approximation error of a continuous function. Furthermore, to advance CVNNs in the field of wireless communications, this paper focuses on the transformer model, which represents a more sophisticated deep learning architecture and has been shown to have excellent performance in wireless communications but only in its real-valued form. In this aspect, we propose a fundamental paradigm of complex-valued transformers for wireless communications. Leveraging this structure, we develop customized complex-valued transformers for three representative applications in wireless communications: channel estimation, user activity detection, and precoding design. These applications utilize transformers with varying levels of sophistication and span a variety of tasks, ranging from regression to classification, supervised to unsupervised learning, and specific module design to end-to-end design. Experimental results demonstrate the superior performance of the complex-valued transformers for the above three applications compared to other traditional real-valued neural network-based methods.
This article introduces DT4ECG, an innovative dual-task learning framework for Electrocardiogram (ECG)-based human identity recognition and activity detection. The framework employs a robust one-dimensional convolutional neural network (1D-CNN) backbone integrated with residual blocks to extract discriminative ECG features. To enhance feature representation, we propose a novel Sequence Channel Attention (SCA) mechanism, which combines channel-wise and sequential context attention to prioritize informative features across both temporal and channel dimensions. Furthermore, to address gradient imbalance in multi-task learning, we integrate GradNorm, a technique that dynamically adjusts loss weights based on gradient magnitudes, ensuring balanced training across tasks. Experimental results demonstrate the superior performance of our model, achieving accuracy rates of 99.12% in ID classification and 90.11% in activity classification. These findings underscore the potential of the DT4ECG framework in enhancing security and user experience across various applications such as fitness monitoring and personalized healthcare, thereby presenting a transformative approach to integrating ECG-based biometrics in everyday technologies.
In this paper, we introduce a multi-talker distant automatic speech recognition (DASR) system we designed for the DASR task 1 of the CHiME-8 challenge. Our system performs speaker counting, diarization, and ASR. It handles various recording conditions, from diner parties to professional meetings and from two to eight speakers. We perform diarization first, followed by speech enhancement, and then ASR as the challenge baseline. However, we introduced several key refinements. First, we derived a powerful speaker diarization relying on end-to-end speaker diarization with vector clustering (EEND-VC), multi-channel speaker counting using enhanced embeddings from EEND-VC, and target-speaker voice activity detection (TS-VAD). For speech enhancement, we introduced a novel microphone selection rule to better select the most relevant microphones among the distributed microphones and investigated improvements to beamforming. Finally, for ASR, we developed several models exploiting Whisper and WavLM speech foundation models. We present the results we submitted to the challenge and updated results we obtained afterward. Our strongest system achieves a 63% relative macro tcpWER improvement over the baseline and outperforms the challenge best results on the NOTSOFAR-1 meeting evaluation data among geometry-independent systems.
Purpose: Surgical performance depends not only on surgeons' technical skills but also on team communication within and across the different professional groups present during the operation. Therefore, automatically identifying team communication in the OR is crucial for patient safety and advances in the development of computer-assisted surgical workflow analysis and intra-operative support systems. To take the first step, we propose a new task of detecting communication briefings involving all OR team members, i.e. the team Time-out and the StOP?-protocol, by localizing their start and end times in video recordings of surgical operations. Methods: We generate an OR dataset of real surgeries, called Team-OR, with more than one hundred hours of surgical videos captured by the multi-view camera system in the OR. The dataset contains temporal annotations of 33 Time-out and 22 StOP?-protocol activities in total. We then propose a novel group activity detection approach, where we encode both scene context and action features, and use an efficient neural network model to output the results. Results: The experimental results on the Team-OR dataset show that our approach outperforms existing state-of-the-art temporal action detection approaches. It also demonstrates the lack of research on group activities in the OR, proving the significance of our dataset. Conclusion: We investigate the Team Time-Out and the StOP?-protocol in the OR, by presenting the first OR dataset with temporal annotations of group activities protocols, and introducing a novel group activity detection approach that outperforms existing approaches. Code is available at https://github.com/CAMMA-public/Team-OR.
The spatial diversity and multiplexing advantages of massive multi-input-multi-output (mMIMO) can significantly improve the capacity of massive non-orthogonal multiple access (NOMA) in machine type communications. However, state-of-the-art grant-free massive NOMA schemes for mMIMO systems require accurate estimation of random access channels to perform activity detection and the following coherent data demodulation, which suffers from excessive pilot overhead and access latency. To address this, we propose a pre-equalization aided grant-free massive access scheme for mMIMO systems, where an iterative detection scheme is conceived. Specifically, the base station (BS) firstly activates one of its antennas (i.e., beacon antenna) to broadcast a beacon signal, which facilitates the user equipment (UEs) to perform downlink channel estimation and pre-equalize the uplink random access signal with respect to the channels associated with the beacon antenna. During the uplink transmission stage, the BS detects UEs' activity and data by using the proposed iterative detection algorithm, which consists of three modules: coarse data detection (DD), data-aided channel estimation (CE), and fine DD. In the proposed algorithm, the joint activity and DD is firstly performed based on the signals received by the beacon antenna. Subsequently, the DD is further refined by iteratively performing data-aided CE module and fine DD module using signals received by all BS antennas. Our simulation results demonstrate that the proposed scheme outperforms state-of-the-art mMIMO-based grant-free massive NOMA schemes with the same access latency. Simulation codes are provided to reproduce the results in this article: https://github.com/owenwang517/tvt-2025.
Neovascular age-related macular degeneration (nAMD) is a leading cause of vision loss among older adults, where disease activity detection and progression prediction are critical for nAMD management in terms of timely drug administration and improving patient outcomes. Recent advancements in deep learning offer a promising solution for predicting changes in AMD from optical coherence tomography (OCT) retinal volumes. In this work, we proposed deep learning models for the two tasks of the public MARIO Challenge at MICCAI 2024, designed to detect and forecast changes in nAMD severity with longitudinal retinal OCT. For the first task, we employ a Vision Transformer (ViT) based Siamese Network to detect changes in AMD severity by comparing scan embeddings of a patient from different time points. To train a model to forecast the change after 3 months, we exploit, for the first time, an Earth Mover (Wasserstein) Distance-based loss to harness the ordinal relation within the severity change classes. Both models ranked high on the preliminary leaderboard, demonstrating that their predictive capabilities could facilitate nAMD treatment management.
Determining 'who spoke what and when' remains challenging in real-world applications. In typical scenarios, Speaker Diarization (SD) is employed to address the problem of 'who spoke when,' while Target Speaker Extraction (TSE) or Target Speaker Automatic Speech Recognition (TSASR) techniques are utilized to resolve the issue of 'who spoke what.' Although some works have achieved promising results by combining SD and TSE systems, inconsistencies remain between SD and TSE regarding both output inconsistency and scenario mismatch. To address these limitations, we propose a Universal Speaker Embedding Free Target Speaker Extraction and Personal Voice Activity Detection (USEF-TP) model that jointly performs TSE and Personal Voice Activity Detection (PVAD). USEF-TP leverages frame-level features obtained through a cross-attention mechanism as speaker-related features instead of using speaker embeddings as in traditional approaches. Additionally, a multi-task learning algorithm with a scenario-aware differentiated loss function is applied to ensure robust performance across various levels of speaker overlap. The experimental results show that our proposed USEF-TP model achieves superior performance in TSE and PVAD tasks on the LibriMix and SparseLibriMix datasets. The results on the CALLHOME dataset demonstrate the competitive performance of our model on real recordings.
Target-Speaker Voice Activity Detection (TS-VAD) is the task of detecting the presence of speech from a known target-speaker in an audio frame. Recently, deep neural network-based models have shown good performance in this task. However, training these models requires extensive labelled data, which is costly and time-consuming to obtain, particularly if generalization to unseen environments is crucial. To mitigate this, we propose a causal, Self-Supervised Learning (SSL) pretraining framework, called Denoising Autoregressive Predictive Coding (DN-APC), to enhance TS-VAD performance in noisy conditions. We also explore various speaker conditioning methods and evaluate their performance under different noisy conditions. Our experiments show that DN-APC improves performance in noisy conditions, with a general improvement of approx. 2% in both seen and unseen noise. Additionally, we find that FiLM conditioning provides the best overall performance. Representation analysis via tSNE plots reveals robust initial representations of speech and non-speech from pretraining. This underscores the effectiveness of SSL pretraining in improving the robustness and performance of TS-VAD models in noisy environments.
This paper sets out the first web-based transcription system for the Irish language - Fotheidil, a system that utilises speech-related AI technologies as part of the ABAIR initiative. The system includes both off-the-shelf pre-trained voice activity detection and speaker diarisation models and models trained specifically for Irish automatic speech recognition and capitalisation and punctuation restoration. Semi-supervised learning is explored to improve the acoustic model of a modular TDNN-HMM ASR system, yielding substantial improvements for out-of-domain test sets and dialects that are underrepresented in the supervised training set. A novel approach to capitalisation and punctuation restoration involving sequence-to-sequence models is compared with the conventional approach using a classification model. Experimental results show here also substantial improvements in performance. The system will be made freely available for public use, and represents an important resource to researchers and others who transcribe Irish language materials. Human-corrected transcriptions will be collected and included in the training dataset as the system is used, which should lead to incremental improvements to the ASR model in a cyclical, community-driven fashion.
We address the challenge of WiFi-based temporal activity detection and propose an efficient Dual Pyramid Network that integrates Temporal Signal Semantic Encoders and Local Sensitive Response Encoders. The Temporal Signal Semantic Encoder splits feature learning into high and low-frequency components, using a novel Signed Mask-Attention mechanism to emphasize important areas and downplay unimportant ones, with the features fused using ContraNorm. The Local Sensitive Response Encoder captures fluctuations without learning. These feature pyramids are then combined using a new cross-attention fusion mechanism. We also introduce a dataset with over 2,114 activity segments across 553 WiFi CSI samples, each lasting around 85 seconds. Extensive experiments show our method outperforms challenging baselines.
The classification of harmful brain activities, such as seizures and periodic discharges, play a vital role in neurocritical care, enabling timely diagnosis and intervention. Electroencephalography (EEG) provides a non-invasive method for monitoring brain activity, but the manual interpretation of EEG signals are time-consuming and rely heavily on expert judgment. This study presents a comparative analysis of deep learning architectures, including Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and EEGNet, applied to the classification of harmful brain activities using both raw EEG data and time-frequency representations generated through Continuous Wavelet Transform (CWT). We evaluate the performance of these models use multimodal data representations, including high-resolution spectrograms and waveform data, and introduce a multi-stage training strategy to improve model robustness. Our results show that training strategies, data preprocessing, and augmentation techniques are as critical to model success as architecture choice, with multi-stage TinyViT and EfficientNet demonstrating superior performance. The findings underscore the importance of robust training regimes in achieving accurate and efficient EEG classification, providing valuable insights for deploying AI models in clinical practice.
The problem of uplink transmissions in massive connectivity is commonly dealt with using schemes for grant-free random access. When a large number of devices transmit almost synchronously, the receiver may not be able to resolve the collision. This could be addressed by assigning dedicated pilots to each user, leading to a contention-free random access (CFRA), which suffers from low scalability and efficiency. This paper explores contention-based random access (CBRA) schemes for asynchronous access in massive multiple-input multiple-output (MIMO) systems. The symmetry across the accessing users with the same pilots is broken by leveraging the delay information inherent to asynchronous systems and the angle information from massive MIMO to enhance activity detection (AD) and channel estimation (CE). The problem is formulated as a sparse recovery in the delay-angle domain. The challenge is that the recovery signal exhibits both row-sparse and cluster-sparse structure, with unknown cluster sizes and locations. We address this by a cluster-extended sparse Bayesian learning (CE-SBL) algorithm that introduces a new weighted prior to capture the signal structure and extends the expectation maximization (EM) algorithm for hyperparameter estimation. Simulation results demonstrate the superiority of the proposed method in joint AD and CE.
This work introduces the first framework for reconstructing surgical dialogue from unstructured real-world recordings, which is crucial for characterizing teaching tasks. In surgical training, the formative verbal feedback that trainers provide to trainees during live surgeries is crucial for ensuring safety, correcting behavior immediately, and facilitating long-term skill acquisition. However, analyzing and quantifying this feedback is challenging due to its unstructured and specialized nature. Automated systems are essential to manage these complexities at scale, allowing for the creation of structured datasets that enhance feedback analysis and improve surgical education. Our framework integrates voice activity detection, speaker diarization, and automated speech recaognition, with a novel enhancement that 1) removes hallucinations (non-existent utterances generated during speech recognition fueled by noise in the operating room) and 2) separates speech from trainers and trainees using few-shot voice samples. These aspects are vital for reconstructing accurate surgical dialogues and understanding the roles of operating room participants. Using data from 33 real-world surgeries, we demonstrated the system's capability to reconstruct surgical teaching dialogues and detect feedback instances effectively (F1 score of 0.79+/-0.07). Moreover, our hallucination removal step improves feedback detection performance by ~14%. Evaluation on downstream clinically relevant tasks of predicting Behavioral Adjustment of trainees and classifying Technical feedback, showed performances comparable to manual annotations with F1 scores of 0.82+/0.03 and 0.81+/0.03 respectively. These results highlight the effectiveness of our framework in supporting clinically relevant tasks and improving over manual methods.
This paper proposes a novel Sequence-to-Sequence Neural Diarization (S2SND) framework to perform online and offline speaker diarization. It is developed from the sequence-to-sequence architecture of our previous target-speaker voice activity detection system and then evolves into a new diarization paradigm by addressing two critical problems. 1) Speaker Detection: The proposed approach can utilize partially given speaker embeddings to discover the unknown speaker and predict the target voice activities in the audio signal. It does not require a prior diarization system for speaker enrollment in advance. 2) Speaker Representation: The proposed approach can adopt the predicted voice activities as reference information to extract speaker embeddings from the audio signal simultaneously. The representation space of speaker embedding is jointly learned within the whole diarization network without using an extra speaker embedding model. During inference, the S2SND framework can process long audio recordings blockwise. The detection module utilizes the previously obtained speaker-embedding buffer to predict both enrolled and unknown speakers' voice activities for each coming audio block. Next, the speaker-embedding buffer is updated according to the predictions of the representation module. Assuming that up to one new speaker may appear in a small block shift, our model iteratively predicts the results of each block and extracts target embeddings for the subsequent blocks until the signal ends. Finally, the last speaker-embedding buffer can re-score the entire audio, achieving highly accurate diarization performance as an offline system. Experimental results show that ...
We propose a novel generalized framework for grant-free random-access (GFRA) in cell-free massive multiple input multiple-output systems where multiple geographically separated access points (APs) or base stations (BSs) aim to detect sporadically active user-equipment (UEs). Unlike a conventional architecture in which all the active UEs transmit their signature or pilot sequences of equal length, we admit a flexible pilot length for each UE, which also enables a seamless integration into conventional grant-based wireless systems. We formulate the joint UE activity detection and the distributed channel estimation as a sparse support and signal recovery problem, and describe a Bayesian learning procedure to solve it. We develop a scheme to fuse the posterior statistics of the latent variables inferred by each AP to jointly detect the UEs' activities, and utilize them to further refine the channel estimates. In addition, we allude to an interesting point which enables this flexible GFRA framework to encode the information bits from the active UEs. We numerically evaluate the normalized mean square error and the probability of miss-detection performances obtained by the Bayesian algorithm and show that the latent-variable fusion enhances the detection and the channel estimation performances by a large margin. We also benchmark against a genie-aided algorithm which has a prior knowledge of the UEs' activities.
Given the extensive research and real-world applications of automatic speech recognition (ASR), ensuring the robustness of ASR models against minor input perturbations becomes a crucial consideration for maintaining their effectiveness in real-time scenarios. Previous explorations into ASR model robustness have predominantly revolved around evaluating accuracy on white-box settings with full access to ASR models. Nevertheless, full ASR model details are often not available in real-world applications. Therefore, evaluating the robustness of black-box ASR models is essential for a comprehensive understanding of ASR model resilience. In this regard, we thoroughly study the vulnerability of practical black-box attacks in cutting-edge ASR models and propose to employ two advanced time-domain-based transferable attacks alongside our differentiable feature extractor. We also propose a speech-aware gradient optimization approach (SAGO) for ASR, which forces mistranscription with minimal impact on human imperceptibility through voice activity detection rule and a speech-aware gradient-oriented optimizer. Our comprehensive experimental results reveal performance enhancements compared to baseline approaches across five models on two databases.
This work considers an uplink asynchronous massive random access scenario in which a large number of users asynchronously access a base station equipped with multiple receive antennas. The objective is to alleviate the problem of massive collision due to the limited number of orthogonal preambles of an access scheme in which user activity detection is performed. We propose a user activity detection with delay-calibration (UAD-DC) algorithm and investigate the benefits of oversampling for the estimation of continuous time delays at the receiver. The proposed algorithm iteratively estimates time delays and detects active users by noting that the collided users can be identified through accurate estimation of time delays. Due to the sporadic user activity patterns, the user activity detection problem can be formulated as a compressive sensing (CS) problem, which can be solved by a modified Turbo-CS algorithm under the consideration of correlated noise samples resulting from oversampling. A sliding-window technique is applied in the proposed algorithm to reduce the overall computational complexity. Moreover, we propose a new design of the pulse shaping filter by minimizing the Bayesian Cram\'er-Rao bound of the detection problem under the constraint of limited spectral bandwidth. Numerical results demonstrate the efficacy of the proposed algorithm in terms of the normalized mean squared error of the estimated channel, the probability of misdetection and the successful detection ratio.
In spite of the popularity of end-to-end diarization systems nowadays, modular systems comprised of voice activity detection (VAD), speaker embedding extraction plus clustering, and overlapped speech detection (OSD) plus handling still attain competitive performance in many conditions. However, one of the main drawbacks of modular systems is the need to run (and train) different modules independently. In this work, we propose an approach to jointly train a model to produce speaker embeddings, VAD and OSD simultaneously and reach competitive performance at a fraction of the inference time of a standard approach. Furthermore, the joint inference leads to a simplified overall pipeline which brings us one step closer to a unified clustering-based method that can be trained end-to-end towards a diarization-specific objective.
Surveillance systems often struggle with managing vast amounts of footage, much of which is irrelevant, leading to inefficient storage and challenges in event retrieval. This paper addresses these issues by proposing an optimized video recording solution focused on activity detection. The proposed approach utilizes a hybrid method that combines motion detection via frame subtraction with object detection using YOLOv9. This strategy specifically targets the recording of scenes involving human or car activity, thereby reducing unnecessary footage and optimizing storage usage. The developed model demonstrates superior performance, achieving precision metrics of 0.855 for car detection and 0.884 for person detection, and reducing the storage requirements by two-thirds compared to traditional surveillance systems that rely solely on motion detection. This significant reduction in storage highlights the effectiveness of the proposed approach in enhancing surveillance system efficiency. Nonetheless, some limitations persist, particularly the occurrence of false positives and false negatives in adverse weather conditions, such as strong winds.