PyTorchKR Papers

최신 논문

총 361편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2203.07093

Human Attention Detection Using AM-FM Representations

Human activity detection from digital videos presents many challenges to the computer vision and image processing communities. Recently, many methods have been developed to detect human activities with varying degree of success. Yet, the general human activity detection problem remains very challenging, especially when the methods need to work 'in the wild' (e.g., without having precise control over the imaging geometry). The thesis explores phase-based solutions for (i) detecting faces, (ii) back of the heads, (iii) joint detection of faces and back of the heads, and (iv) whether the head is looking to the left or the right, using standard video cameras without any control on the imaging geometry. The proposed phase-based approach is based on the development of simple and robust methods that rely on the use of Amplitude Modulation- Frequency Modulation (AM-FM) models. The approach is validated using video frames extracted from the Advancing Out-of-school Learning in Mathematics and Engineering (AOLME) project. The dataset consisted of 13,265 images from ten students looking at the camera, and 6,122 images from five students looking away from the camera. For the students facing the camera, the method was able to correctly classify 97.1% of them looking to the left and 95.9% of them looking to the right. For the students facing the back of the camera, the method was able to correctly classify 87.6% of them looking to the left and 93.3% of them looking to the right. The results indicate that AM-FM based methods hold great promise for analyzing human activity videos.

arXiv:2203.03796

PAMI-AD: An Activity Detector Exploiting Part-attention and Motion Information in Surveillance Videos

Activity detection in surveillance videos is a challenging task caused by small objects, complex activity categories, its untrimmed nature, etc. Existing methods are generally limited in performance due to inaccurate proposals, poor classifiers or inadequate post-processing method. In this work, we propose a comprehensive and effective activity detection system in untrimmed surveillance videos for person-centered and vehicle-centered activities. It consists of four modules, i.e., object localizer, proposal filter, activity classifier and activity refiner. For person-centered activities, a novel part-attention mechanism is proposed to explore detailed features in different body parts. As for vehicle-centered activities, we propose a localization masking method to jointly encode motion and foreground attention features. We conduct experiments on the large-scale activity detection datasets VIRAT, and achieve the best results for both groups of activities. Furthermore, our team won the 1st place in the TRECVID 2021 ActEV challenge.

arXiv:2202.13058

Random Access with Massive MIMO-OTFS in LEO Satellite Communications

This paper considers the joint channel estimation and device activity detection in the grant-free random access systems, where a large number of Internet-of-Things devices intend to communicate with a low-earth orbit satellite in a sporadic way. In addition, the massive multiple-input multiple-output (MIMO) with orthogonal time-frequency space (OTFS) modulation is adopted to combat the dynamics of the terrestrial-satellite link. We first analyze the input-output relationship of the single-input single-output OTFS when the large delay and Doppler shift both exist, and then extend it to the grant-free random access with massive MIMO-OTFS. Next, by exploring the sparsity of channel in the delay-Doppler-angle domain, a two-dimensional pattern coupled hierarchical prior with the sparse Bayesian learning and covariance-free method (TDSBL-CF) is developed for the channel estimation. Then, the active devices are detected by computing the energy of the estimated channel. Finally, the generalized approximate message passing algorithm combined with the sparse Bayesian learning and two-dimensional convolution (ConvSBL-GAMP) is proposed to decrease the computations of the TDSBL-CF algorithm. Simulation results demonstrate that the proposed algorithms outperform conventional methods.

arXiv:2202.10593

VADOI:Voice-Activity-Detection Overlapping Inference For End-to-end Long-form Speech Recognition

While end-to-end models have shown great success on the Automatic Speech Recognition task, performance degrades severely when target sentences are long-form. The previous proposed methods, (partial) overlapping inference are shown to be effective on long-form decoding. For both methods, word error rate (WER) decreases monotonically when overlapping percentage decreases. Setting aside computational cost, the setup with 50% overlapping during inference can achieve the best performance. However, a lower overlapping percentage has an advantage of fast inference speed. In this paper, we first conduct comprehensive experiments comparing overlapping inference and partial overlapping inference with various configurations. We then propose Voice-Activity-Detection Overlapping Inference to provide a trade-off between WER and computation cost. Results show that the proposed method can achieve a 20% relative computation cost reduction on Librispeech and Microsoft Speech Language Translation long-form corpus while maintaining the WER performance when comparing to the best performing overlapping inference algorithm. We also propose Soft-Match to compensate for similar words mis-aligned problem.

arXiv:2202.05833

Active Privacy-Utility Trade-off Against Inference in Time-Series Data Sharing

Internet of things (IoT) devices, such as smart meters, smart speakers and activity monitors, have become highly popular thanks to the services they offer. However, in addition to their many benefits, they raise privacy concerns since they share fine-grained time-series user data with untrusted third parties. In this work, we consider a user releasing her data containing personal information in return of a service from an honest-but-curious service provider (SP). We model user's personal information as two correlated random variables (r.v.'s), one of them, called the secret variable, is to be kept private, while the other, called the useful variable, is to be disclosed for utility. We consider active sequential data release, where at each time step the user chooses from among a finite set of release mechanisms, each revealing some information about the user's personal information, i.e., the true values of the r.v.'s, albeit with different statistics. The user manages data release in an online fashion such that the maximum amount of information is revealed about the latent useful variable as quickly as possible, while the confidence for the sensitive variable is kept below a predefined level. For privacy measure, we consider both the probability of correctly detecting the true value of the secret and the mutual information (MI) between the secret and the released data. We formulate both problems as partially observable Markov decision processes (POMDPs), and numerically solve them by advantage actor-critic (A2C) deep reinforcement learning (DRL). We evaluate the privacy-utility trade-off (PUT) of the proposed policies on both the synthetic data and smoking activity dataset, and show their validity by testing the activity detection accuracy of the SP modeled by a long short-term memory (LSTM) neural network.

arXiv:2202.04855

The USTC-Ximalaya system for the ICASSP 2022 multi-channel multi-party meeting transcription (M2MeT) challenge

We propose two improvements to target-speaker voice activity detection (TS-VAD), the core component in our proposed speaker diarization system that was submitted to the 2022 Multi-Channel Multi-Party Meeting Transcription (M2MeT) challenge. These techniques are designed to handle multi-speaker conversations in real-world meeting scenarios with high speaker-overlap ratios and under heavy reverberant and noisy condition. First, for data preparation and augmentation in training TS-VAD models, speech data containing both real meetings and simulated indoor conversations are used. Second, in refining results obtained after TS-VAD based decoding, we perform a series of post-processing steps to improve the VAD results needed to reduce diarization error rates (DERs). Tested on the ALIMEETING corpus, the newly released Mandarin meeting dataset used in M2MeT, we demonstrate that our proposed system can decrease the DER by up to 66.55/60.59% relatively when compared with classical clustering based diarization on the Eval/Test set.

arXiv:2202.02687

Cross-Channel Attention-Based Target Speaker Voice Activity Detection: Experimental Results for M2MeT Challenge

In this paper, we present the speaker diarization system for the Multi-channel Multi-party Meeting Transcription Challenge (M2MeT) from team DKU_DukeECE. As the highly overlapped speech exists in the dataset, we employ an x-vector-based target-speaker voice activity detection (TS-VAD) to find the overlap between speakers. For the single-channel scenario, we separately train a model for each of the 8 channels and fuse the results. We also employ the cross-channel self-attention to further improve the performance, where the non-linear spatial correlations between different channels are learned and fused. Experimental results on the evaluation set show that the single-channel TS-VAD reduces the DER by over 75% from 12.68\% to 3.14%. The multi-channel TS-VAD further reduces the DER by 28% and achieves a DER of 2.26%. Our final submitted system achieves a DER of 2.98% on the AliMeeting test set, which ranks 1st in the M2MET challenge.

arXiv:2202.01986

The CUHK-TENCENT speaker diarization system for the ICASSP 2022 multi-channel multi-party meeting transcription challenge

This paper describes our speaker diarization system submitted to the Multi-channel Multi-party Meeting Transcription (M2MeT) challenge, where Mandarin meeting data were recorded in multi-channel format for diarization and automatic speech recognition (ASR) tasks. In these meeting scenarios, the uncertainty of the speaker number and the high ratio of overlapped speech present great challenges for diarization. Based on the assumption that there is valuable complementary information between acoustic features, spatial-related and speaker-related features, we propose a multi-level feature fusion mechanism based target-speaker voice activity detection (FFM-TS-VAD) system to improve the performance of the conventional TS-VAD system. Furthermore, we propose a data augmentation method during training to improve the system robustness when the angular difference between two speakers is relatively small. We provide comparisons for different sub-systems we used in M2MeT challenge. Our submission is a fusion of several sub-systems and ranks second in the diarization task.

arXiv:2201.12755공식 코드 · Papers with Code

HGCN: Harmonic gated compensation network for speech enhancement

Mask processing in the time-frequency (T-F) domain through the neural network has been one of the mainstreams for single-channel speech enhancement. However, it is hard for most models to handle the situation when harmonics are partially masked by noise. To tackle this challenge, we propose a harmonic gated compensation network (HGCN). We design a high-resolution harmonic integral spectrum to improve the accuracy of harmonic locations prediction. Then we add voice activity detection (VAD) and voiced region detection (VRD) to the convolutional recurrent network (CRN) to filter harmonic locations. Finally, the harmonic gating mechanism is used to guide the compensation model to adjust the coarse results from CRN to obtain the refinedly enhanced results. Our experiments show HGCN achieves substantial gain over a number of advanced approaches in the community.

arXiv:2201.09032공식 코드 · Papers with Code

NAS-VAD: Neural Architecture Search for Voice Activity Detection

Various neural network-based approaches have been proposed for more robust and accurate voice activity detection (VAD). Manual design of such neural architectures is an error-prone and time-consuming process, which prompted the development of neural architecture search (NAS) that automatically design and optimize network architectures. While NAS has been successfully applied to improve performance in a variety of tasks, it has not yet been exploited in the VAD domain. In this paper, we present the first work that utilizes NAS approaches on the VAD task. To effectively search architectures for the VAD task, we propose a modified macro structure and a new search space with a much broader range of operations that includes attention operations. The results show that the network structures found by the propose NAS framework outperform previous manually designed state-of-the-art VAD models in various noise-added and real-world-recorded datasets. We also show that the architectures searched on a particular dataset achieve improved generalization performance on unseen audio datasets. Our code and models are available at https://github.com/daniel03c1/NAS_VAD.

arXiv:2201.05290

Argus++: Robust Real-time Activity Detection for Unconstrained Video Streams with Overlapping Cube Proposals

Activity detection is one of the attractive computer vision tasks to exploit the video streams captured by widely installed cameras. Although achieving impressive performance, conventional activity detection algorithms are usually designed under certain constraints, such as using trimmed and/or object-centered video clips as inputs. Therefore, they failed to deal with the multi-scale multi-instance cases in real-world unconstrained video streams, which are untrimmed and have large field-of-views. Real-time requirements for streaming analysis also mark brute force expansion of them unfeasible. To overcome these issues, we propose Argus++, a robust real-time activity detection system for analyzing unconstrained video streams. The design of Argus++ introduces overlapping spatio-temporal cubes as an intermediate concept of activity proposals to ensure coverage and completeness of activity detection through over-sampling. The overall system is optimized for real-time processing on standalone consumer-level hardware. Extensive experiments on different surveillance and driving scenarios demonstrated its superior performance in a series of activity detection benchmarks, including CVPR ActivityNet ActEV 2021, NIST ActEV SDL UF/KF, TRECVID ActEV 2020/2021, and ICCV ROAD 2021.

arXiv:2201.01928

Egocentric Deep Multi-Channel Audio-Visual Active Speaker Localization

Augmented reality devices have the potential to enhance human perception and enable other assistive functionalities in complex conversational environments. Effectively capturing the audio-visual context necessary for understanding these social interactions first requires detecting and localizing the voice activities of the device wearer and the surrounding people. These tasks are challenging due to their egocentric nature: the wearer's head motion may cause motion blur, surrounding people may appear in difficult viewing angles, and there may be occlusions, visual clutter, audio noise, and bad lighting. Under these conditions, previous state-of-the-art active speaker detection methods do not give satisfactory results. Instead, we tackle the problem from a new setting using both video and multi-channel microphone array audio. We propose a novel end-to-end deep learning approach that is able to give robust voice activity detection and localization results. In contrast to previous methods, our method localizes active speakers from all possible directions on the sphere, even outside the camera's field of view, while simultaneously detecting the device wearer's own voice activity. Our experiments show that the proposed method gives superior results, can run in real time, and is robust against noise and clutter.

arXiv:2201.01485

Exploiting Temporal Side Information in Massive IoT Connectivity

This paper considers the joint device activity detection and channel estimation problem in a massive Internet of Things (IoT) connectivity system, where a large number of IoT devices exist but merely a random subset of them become active for short-packet transmission in each coherence block. In particular, we propose to leverage the temporal correlation in device activity, e.g., a device active in the previous coherence block is more likely to be still active in the current coherence block, to improve the detection and estimation performance. However, it is challenging to utilize this temporal correlation as side information (SI), which relies on the knowledge about the exact statistical relation between the estimated activity pattern for the previous coherence block (which may be imperfect with unknown error) and the true activity pattern in the current coherence block. To tackle this challenge, we establish a novel SI-aided multiple measurement vector approximate message passing (MMV-AMP) framework. Specifically, thanks to the state evolution of the MMV-AMP algorithm, the correlation between the activity pattern estimated by the MMV-AMP algorithm in the previous coherence block and the real activity pattern in the current coherence block is quantified explicitly. Based on the well-defined temporal correlation, we further manage to embed this useful SI into the denoiser design under the MMV-AMP framework. Specifically, the SI-based soft-thresholding denoisers with binary thresholds and the SI-based minimum mean-squared error (MMSE) denoisers are characterized for the cases without and with the knowledge of the channel distribution, respectively. Numerical results are given to show the significant gain in device activity detection and channel estimation performance brought by our proposed SI-aided MMV-AMP framework.

arXiv:2112.11824

Binary Image Skeletonization Using 2-Stage U-Net

Object Skeletonization is the process of extracting skeletal, line-like representations of shapes. It provides a very useful tool for geometric shape understanding and minimal shape representation. It also has a wide variety of applications, most notably in anatomical research and activity detection. Several mathematical algorithmic approaches have been developed to solve this problem, and some of them have been proven quite robust. However, a lesser amount of attention has been invested into deep learning solutions for it. In this paper, we use a 2-stage variant of the famous U-Net architecture to split the problem space into two sub-problems: shape minimization and corrective skeleton thinning. Our model produces results that are visually much better than the baseline SkelNetOn model. We propose a new metric, M-CCORR, based on normalized correlation coefficients as an alternative to F1 for this challenge as it solves the problem of class imbalance, managing to recognize skeleton similarity without suffering from F1's over-sensitivity to pixel-shifts.

arXiv:2112.09042

Low Resource Species Agnostic Bird Activity Detection

This paper explores low resource classifiers and features for the detection of bird activity, suitable for embedded Automatic Recording Units which are typically deployed for long term remote monitoring of bird populations. Features include low-level spectral parameters, statistical moments on pitch samples, and features derived from amplitude modulation. Performance is evaluated on several lightweight classifiers using the NIPS4Bplus dataset. Our experiments show that random forest classifiers perform best on this task, achieving an accuracy of 0.721 and an F1-Score of 0.604. We compare the results of our system against both a Convolutional Neural Network based detector, and standard MFCC features. Our experiments show that we can achieve equal or better performance in most metrics using features and models with a smaller computational cost and which are suitable for edge deployment.

arXiv:2112.05016공식 코드 · Papers with Code

X-Vector based voice activity detection for multi-genre broadcast speech-to-text

Voice Activity Detection (VAD) is a fundamental preprocessing step in automatic speech recognition. This is especially true within the broadcast industry where a wide variety of audio materials and recording conditions are encountered. Based on previous studies which indicate that xvector embeddings can be applied to a diverse set of audio classification tasks, we investigate the suitability of x-vectors in discriminating speech from noise. We find that the proposed x-vector based VAD system achieves the best reported score in detecting clean speech on AVA-Speech, whilst retaining robust VAD performance in the presence of noise and music. Furthermore, we integrate the x-vector based VAD system into an existing STT pipeline and compare its performance on multiple broadcast datasets against a baseline system with WebRTC VAD. Crucially, our proposed x-vector based VAD improves the accuracy of STT transcription on real-world broadcast audio

arXiv:2112.04295

User Activity Detection and Channel Estimation of Spatially Correlated Channels via AMP in Massive MTC

This paper addresses the problem of joint user identification and channel estimation (JUICE) for grant-free access in massive machine-type communications (mMTC). We consider the JUICE under a spatially correlated fading channel model as that reflects the main characteristics of the practical multiple-input multiple-output channels. We formulate the JUICE as a sparse recovery problem in a multiple measurement vector setup and present a solution based on the approximate message passing (AMP) algorithm that takes into account the channel spatial correlation. Using the state evolution, we provide a detailed theoretical analysis on the activity detection performance of AMP-based JUICE by deriving closed-from expressions for the probabilities of miss detection and false alarm. The simulation experiments show that the performance predicted by the theoretical analysis matches the one obtained by the numerical results.

arXiv:2112.02830

Learning Proximal Operator Methods for Massive Connectivity in IoT Networks

Grant-free random access has the potential to support massive connectivity in Internet of Things (IoT) networks, where joint activity detection and channel estimation (JADCE) is a key issue that needs to be tackled. The existing methods for JADCE usually suffer from one of the following limitations: high computational complexity, ineffective in inducing sparsity, and incapable of handling complex matrix estimation. To mitigate all the aforementioned limitations, we in this paper develop an effective unfolding neural network framework built upon the proximal operator method to tackle the JADCE problem in IoT networks, where the base station is equipped with multiple antennas. Specifically, the JADCE problem is formulated as a group-sparse-matrix estimation problem, which is regularized by non-convex minimax concave penalty (MCP). This problem can be iteratively solved by using the proximal operator method, based on which we develop a unfolding neural network structure by parameterizing the algorithmic iterations. By further exploiting the coupling structure among the training parameters as well as the analytical computation, we develop two additional unfolding structures to reduce the training complexity. We prove that the proposed algorithm achieves a linear convergence rate. Results show that our proposed three unfolding structures not only achieve a faster convergence rate but also obtain a higher estimation accuracy than the baseline methods.

arXiv:2111.13694공식 코드 · Papers with Code

Speaker Embedding-aware Neural Diarization for Flexible Number of Speakers with Textual Information

Overlapping speech diarization is always treated as a multi-label classification problem. In this paper, we reformulate this task as a single-label prediction problem by encoding the multi-speaker labels with power set. Specifically, we propose the speaker embedding-aware neural diarization (SEND) method, which predicts the power set encoded labels according to the similarities between speech features and given speaker embeddings. Our method is further extended and integrated with downstream tasks by utilizing the textual information, which has not been well studied in previous literature. The experimental results show that our method achieves lower diarization error rate than the target-speaker voice activity detection. When textual information is involved, the diarization errors can be further reduced. For the real meeting scenario, our method can achieve 34.11% relative improvement compared with the Bayesian hidden Markov model based clustering algorithm.

arXiv:2111.13675공식 코드 · Papers with Code

Weakly-guided Self-supervised Pretraining for Temporal Activity Detection

Temporal Activity Detection aims to predict activity classes per frame, in contrast to video-level predictions in Activity Classification (i.e., Activity Recognition). Due to the expensive frame-level annotations required for detection, the scale of detection datasets is limited. Thus, commonly, previous work on temporal activity detection resorts to fine-tuning a classification model pretrained on large-scale classification datasets (e.g., Kinetics-400). However, such pretrained models are not ideal for downstream detection, due to the disparity between the pretraining and the downstream fine-tuning tasks. In this work, we propose a novel 'weakly-guided self-supervised' pretraining method for detection. We leverage weak labels (classification) to introduce a self-supervised pretext task (detection) by generating frame-level pseudo labels, multi-action frames, and action segments. Simply put, we design a detection task similar to downstream, on large-scale classification data, without extra annotations. We show that the models pretrained with the proposed weakly-guided self-supervised detection task outperform prior work on multiple challenging activity detection benchmarks, including Charades and MultiTHUMOS. Our extensive ablations further provide insights on when and how to use the proposed models for activity detection. Code is available at https://github.com/kkahatapitiya/SSDet.