Every day, humans perform many closely related activities that involve subtle discriminative motions, such as putting on a shirt vs. putting on a jacket, or shaking hands vs. giving a high five. Activity recognition by ethical visual AI could provide insights into our patterns of daily life, however existing activity recognition datasets do not capture the massive diversity of these human activities around the world. To address this limitation, we introduce Collector, a free mobile app to record video while simultaneously annotating objects and activities of consented subjects. This new data collection platform was used to curate the Consented Activities of People (CAP) dataset, the first large-scale, fine-grained activity dataset of people worldwide. The CAP dataset contains 1.45M video clips of 512 fine grained activity labels of daily life, collected by 780 subjects in 33 countries. We provide activity classification and activity detection benchmarks for this dataset, and analyze baseline results to gain insight into how people around with world perform common activities. The dataset, benchmarks, evaluation tools, public leaderboards and mobile apps are available for use at visym.github.io/cap.
Self-supervised-learning-based pre-trained models for speech data, such as Wav2Vec 2.0 (W2V2), have become the backbone of many speech tasks. In this paper, to achieve speaker diarisation and speech recognition using a single model, a tandem multitask training (TMT) method is proposed to fine-tune W2V2. For speaker diarisation, the tasks of voice activity detection (VAD) and speaker classification (SC) are required, and connectionist temporal classification (CTC) is used for ASR. The multitask framework implements VAD, SC, and ASR using an early layer, middle layer, and late layer of W2V2, which coincides with the order of segmenting the audio with VAD, clustering the segments based on speaker embeddings, and transcribing each segment with ASR. Experimental results on the augmented multi-party (AMI) dataset showed that using different W2V2 layers for VAD, SC, and ASR from the earlier to later layers for TMT not only saves computational cost, but also reduces diarisation error rates (DERs). Joint fine-tuning of VAD, SC, and ASR yielded 16%/17% relative reductions of DER with manual/automatic segmentation respectively, and consistent reductions in speaker attributed word error rate, compared to the baseline with separately fine-tuned models.
Voice Activity Detection (VAD) is an important pre-processing step in a wide variety of speech processing systems. VAD should in a practical application be able to detect speech in both noisy and noise-free environments, while not introducing significant latency. In this work we propose using an adversarial multi-task learning method when training a supervised VAD. The method has been applied to the state-of-the-art VAD Waveform-based Voice Activity Detection. Additionally the performance of the VADis investigated under different algorithmic delays, which is an important factor in latency. Introducing adversarial multi-task learning to the model is observed to increase performance in terms of Area Under Curve (AUC), particularly in noisy environments, while the performance is not degraded at higher SNR levels. The adversarial multi-task learning is only applied in the training phase and thus introduces no additional cost in testing. Furthermore the correlation between performance and algorithmic delays is investigated, and it is observed that the VAD performance degradation is only moderate when lowering the algorithmic delay from 398 ms to 23 ms.
An autonomous Artificial Internet of Things (AIoT) system for elderly dementia patients monitoring in a smart home is presented. The system mainly implements two functions based on the activity inference of the sensor data, which are real time abnormal activity monitoring and trend prediction of disease related activities. Specifically, CASAS dataset is employed to train a Random Forest (RF) model for activity inference. Then, another RF model trained by the output data of activity inference is used for abnormal activity monitoring. Particularly, RF is chosen for these tasks because of its balanced trade offs between accuracy, time efficiency, flexibility, and interpretability. Moreover, Long Short Term Memory (LSTM) is utilised to forecast the disease related activity trend of a patient. Consequently, the accuracy of two RF classifiers designed for activity inference and abnormal activity detection is greater than 99 percent and 94 percent, respectively. Furthermore, using the duration of sleep as an example, the LSTM model achieves accurate and evident future trends prediction.
Voice activity detection (VAD) is an important pre-processing step for speech technology applications. The task consists of deriving segment boundaries of audio signals which contain voicing information. In recent years, it has been shown that voice source and vocal tract system information can be extracted using zero-frequency filtering (ZFF) without making any explicit model assumptions about the speech signal. This paper investigates the potential of zero-frequency filtering for jointly modeling voice source and vocal tract system information, and proposes two approaches for VAD. The first approach demarcates voiced regions using a composite signal composed of different zero-frequency filtered signals. The second approach feeds the composite signal as input to the rVAD algorithm. These approaches are compared with other supervised and unsupervised VAD methods in the literature, and are evaluated on the Aurora-2 database, across a range of SNRs (20 to -5 dB). Our studies show that the proposed ZFF-based methods perform comparable to state-of-art VAD methods and are more invariant to added degradation and different channel characteristics.
Speech enhancement and source localization has been active research for several decades with a wide range of real-world applications. Recently, the Deep Complex Convolution Recurrent network (DCCRN) has yielded impressive enhancement performance for single-channel systems. In this study, a neural beamformer consisting of a beamformer and a novel multi-channel DCCRN is proposed for speech enhancement and source localization. Complex-valued filters estimated by the multi-channel DCCRN serve as the weights of beamformer. In addition, a one-stage learning-based procedure is employed for speech enhancement and source localization. The proposed network composed of the multi-channel DCCRN and the auxiliary network models the sound field, while minimizing the distortionless response loss function. Simulation results show that the proposed neural beamformer is effective in enhancing speech signals, with speech quality well preserved. The proposed neural beamformer also provides source localization and voice activity detection (VAD) functions.
The common target speech separation directly estimate the target source, ignoring the interrelationship between different speakers at each frame. We propose a multiple-target speech separation model (MTSS) to simultaneously extract each speaker's voice from the mixed speech rather than just optimally estimating the target source. Moreover, we propose a speaker diarization (SD) aware MTSS system (SD-MTSS), which consists of a SD module and MTSS module. By exploiting the TSVAD decision and the estimated mask, our SD-MTSS model can extract the speech signal of each speaker concurrently in a conversational recording without additional enrollment audio in advance. Experimental results show that our MTSS model achieves 1.38dB SDR, 1.34dB SI-SDR, and 0.13 PESQ improvements over the baseline on the WSJ0-2mix-extr dataset, respectively. The SD-MTSS system makes 19.2% relative speaker dependent character error rate (CER) reduction on the Alimeeting dataset.
This paper studies reconfigurable intelligent surface (RIS) assisted device activity detection for grant-free (GF) uplink transmission in wireless communication networks. In particular, we consider mobile devices located in an area where the direct link to an access point (AP) is blocked. Thus, the devices try to connect to the AP via a reflected link provided by an RIS. Therefore, a RIS phase-shift design is desired that covers the entire blocked area with a wide reflection beam because the exact locations and times of activity of the devices are unknown in GF transmission. In order to study the impact of the phase-shift design on the device activity detection, we derive a generalized likelihood ratio test (GLRT) based detector and present an analytical expression for the probability of detection. Assuming knowledge of statistical CSI, we formulate an optimization problem for the phase-shift design for maximization of the guaranteed probability of detection for all locations within a given coverage area. To tackle the non-convexity of the problem, we propose two different approximations of the objective function. The first approximation leads to a design that aims to reduce the variations of the end-to-end channel while taking system parameters such as transmit power, noise power, and probability of false alarm into account. The second approximation can be adopted for versatile RIS deployments because it only depends on the line-of-sight component of the end-to-end channel and is not affected by system parameters. For comparison, we also consider a phase-shift design maximizing the average channel gain and a baseline analytical phase-shift design for large blocked areas. Our performance evaluation shows that the proposed approximations result in phase-shift designs that guarantee high probability of detection across the coverage area and outperform the baseline designs.
In grant-free sparse code multiple access (GF-SCMA) system, active user detection (AUD) is a major performance bottleneck as it involves complex combinatorial problem, which makes joint design of contention resources for users and AUD at the receiver a crucial but a challenging problem. To this end, we propose autoencoder (AE)-based joint optimization of both preamble generation networks (PGNs) in the encoder side and data-aided AUD in the decoder side. The core architecture of the proposed AE is a novel user activity extraction network (UAEN) in the decoder that extracts a priori user activity information from the SCMA codeword data for the data-aided AUD. An end-to-end training of the proposed AE enables joint optimization of the contention resources, i.e., preamble sequences, each associated with one of the codebooks, and extraction of user activity information from both preamble and SCMA-based data transmission. Furthermore, we propose a self-supervised pre-training scheme for the UAEN prior to the end-to-end training, to ensure the convergence of the UAEN which lies deep inside the AE network. Simulation results demonstrated that the proposed AUD scheme achieved 3 to 5dB gain at a target activity detection error rate of $\bf{{10}^{-3}}$ compared to the state-of-the-art DL-based AUD schemes.
Positive-unlabeled (PU) learning deals with binary classification problems when only positive (P) and unlabeled (U) data are available. Many recent PU methods are based on neural networks, but little has been done to develop boosting algorithms for PU learning, despite boosting algorithms' strong performance on many fully supervised classification problems. In this paper, we propose a novel boosting algorithm, AdaPU, for PU learning. Similarly to AdaBoost, AdaPU aims to optimize an empirical exponential loss, but the loss is based on the PU data, rather than on positive-negative (PN) data. As in AdaBoost, we learn a weighted combination of weak classifiers by learning one weak classifier and its weight at a time. However, AdaPU requires a very different algorithm for learning the weak classifiers and determining their weights. This is because AdaPU learns a weak classifier and its weight using a weighted positive-negative (PN) dataset with some negative data weights $-$ the dataset is derived from the original PU data, and the data weights are determined by the current weighted classifier combination, but some data weights are negative. Our experiments showed that AdaPU outperforms neural networks on several benchmark PU datasets, including a large-scale challenging cyber security dataset.
This letter considers temporal-correlated massive access, where each device, once activated, is likely to transmit continuously over several consecutive frames. Motivated by that the device activity at each frame is correlated to not only its previous frame but also its next frame, we propose a double-sided information (DSI) aided joint activity detection and channel estimation algorithm based on the approximate message passing (AMP) framework. The DSI is extracted from the estimation results in a sliding window that contains the target detection frame and its previous and next frames. The proposed algorithm demonstrates superior performance over the state-of-the-art methods.
Human limb motion tracking and recognition plays an important role in medical rehabilitation training, lower limb assistance, prosthetics design for amputees, feedback control for assistive robots, etc. Lightweight wearable sensors, including inertial sensors, surface electromyography sensors, and flexible strain/pressure, are promising to become the next-generation human motion capture devices. Herein, we present a wireless wearable device consisting of a sixteen-channel flexible sponge-based pressure sensor array to recognize various human lower limb motions by detecting contours on the human skin caused by calf gastrocnemius muscle actions. Each sensing element is a round porous structure of thin carbon nanotube/polydimethylsiloxane nanocomposites with a diameter of 4 mm and thickness of about 400 {\mu}m. Ten human subjects were recruited to perform ten different lower limb motions while wearing the developed device. The motion classification result with the support vector machine method shows a macro-recall of about 97.3% for all ten motions tested. This work demonstrates a portable wearable muscle activity detection device with a lower limb motion recognition application, which can be potentially used in assistive robot control, healthcare, sports monitoring, etc.
Activity recognition in surgical videos is a key research area for developing next-generation devices and workflow monitoring systems. Since surgeries are long processes with highly-variable lengths, deep learning models used for surgical videos often consist of a two-stage setup using a backbone and temporal sequence model. In this paper, we investigate many state-of-the-art backbones and temporal models to find architectures that yield the strongest performance for surgical activity recognition. We first benchmark the models performance on a large-scale activity recognition dataset containing over 800 surgery videos captured in multiple clinical operating rooms. We further evaluate the models on the two smaller public datasets, the Cholec80 and Cataract-101 datasets, containing only 80 and 101 videos respectively. We empirically found that Swin-Transformer+BiGRU temporal model yielded strong performance on both datasets. Finally, we investigate the adaptability of the model to new domains by fine-tuning models to a new hospital and experimenting with a recent unsupervised domain adaptation approach.
We have developed an unsupervised anomalous sound detection method for machine condition monitoring that utilizes an auxiliary task -- detecting when the target machine is active. First, we train a model that detects machine activity by using normal data with machine activity labels and then use the activity-detection error as the anomaly score for a given sound clip if we have access to the ground-truth activity labels in the inference phase. If these labels are not available, the anomaly score is calculated through outlier detection on the embedding vectors obtained by the activity-detection model. Solving this auxiliary task enables the model to learn the difference between the target machine sounds and similar background noise, which makes it possible to identify small deviations in the target sounds. Experimental results showed that the proposed method improves the anomaly-detection performance of the conventional method complementarily by means of an ensemble.
Many archival recordings of speech from endangered languages remain unannotated and inaccessible to community members and language learning programs. One bottleneck is the time-intensive nature of annotation. An even narrower bottleneck occurs for recordings with access constraints, such as language that must be vetted or filtered by authorised community members before annotation can begin. We propose a privacy-preserving workflow to widen both bottlenecks for recordings where speech in the endangered language is intermixed with a more widely-used language such as English for meta-linguistic commentary and questions (e.g. What is the word for 'tree'?). We integrate voice activity detection (VAD), spoken language identification (SLI), and automatic speech recognition (ASR) to transcribe the metalinguistic content, which an authorised person can quickly scan to triage recordings that can be annotated by people with lower levels of access. We report work-in-progress processing 136 hours archival audio containing a mix of English and Muruwari. Our collaborative work with the Muruwari custodian of the archival materials show that this workflow reduces metalanguage transcription time by 20% even given only minimal amounts of annotated training data: 10 utterances per language for SLI and for ASR at most 39 minutes, and possibly as little as 39 seconds.
Personalization of on-device speech recognition (ASR) has seen explosive growth in recent years, largely due to the increasing popularity of personal assistant features on mobile devices and smart home speakers. In this work, we present Personal VAD 2.0, a personalized voice activity detector that detects the voice activity of a target speaker, as part of a streaming on-device ASR system. Although previous proof-of-concept studies have validated the effectiveness of Personal VAD, there are still several critical challenges to address before this model can be used in production: first, the quality must be satisfactory in both enrollment and enrollment-less scenarios; second, it should operate in a streaming fashion; and finally, the model size should be small enough to fit a limited latency and CPU/Memory budget. To meet the multi-faceted requirements, we propose a series of novel designs: 1) advanced speaker embedding modulation methods; 2) a new training paradigm to generalize to enrollment-less conditions; 3) architecture and runtime optimizations for latency and resource restrictions. Extensive experiments on a realistic speech recognition system demonstrated the state-of-the-art performance of our proposed method.
In this paper, we carry out an analysis on the use of speech separation guided diarization (SSGD) in telephone conversations. SSGD performs diarization by separating the speakers signals and then applying voice activity detection on each estimated speaker signal. In particular, we compare two low-latency speech separation models. Moreover, we show a post-processing algorithm that significantly reduces the false alarm errors of a SSGD pipeline. We perform our experiments on two datasets: Fisher Corpus Part 1 and CALLHOME, evaluating both separation and diarization metrics. Notably, our SSGD DPRNN-based online model achieves 11.1% DER on CALLHOME, comparable with most state-of-the-art end-to-end neural diarization models despite being trained on an order of magnitude less data and having considerably lower latency, i.e., 0.1 vs. 10 seconds. We also show that the separated signals can be readily fed to a speech recognition back-end with performance close to the oracle source signals.
Joint activity detection and channel estimation (JADCE) for grant-free random access is a critical issue that needs to be addressed to support massive connectivity in IoT networks. However, the existing model-free learning method can only achieve either activity detection or channel estimation, but not both. In this paper, we propose a novel model-free learning method based on generative adversarial network (GAN) to tackle the JADCE problem. We adopt the U-net architecture to build the generator rather than the standard GAN architecture, where a pre-estimated value that contains the activity information is adopted as input to the generator. By leveraging the properties of the pseudoinverse, the generator is refined by using an affine projection and a skip connection to ensure the output of the generator is consistent with the measurement. Moreover, we build a two-layer fully-connected neural network to design pilot matrix for reducing the impact of receiver noise. Simulation results show that the proposed method outperforms the existing methods in high SNR regimes, as both data consistency projection and pilot matrix optimization improve the learning ability.
Despite the widespread use of encryption techniques to provide confidentiality over Internet communications, mobile device users are still susceptible to privacy and security risks. In this paper, a new Deep Neural Network (DNN) based user activity detection framework is proposed to identify fine grained user activities performed on mobile applications (known as in-app activities) from a sniffed encrypted Internet traffic stream. One of the challenges is that there are countless applications, and it is practically impossible to collect and train a DNN model using all possible data from them. Therefore, in this work we exploit the probability distribution of DNN output layer to filter the data from applications that are not considered during the model training (i.e., unknown data). The proposed framework uses a time window based approach to divide the traffic flow of an activity into segments, so that in-app activities can be identified just by observing only a fraction of the activity related traffic. Our tests have shown that the DNN based framework has demonstrated an accuracy of 90% or above in identifying previously trained in-app activities and an average accuracy of 79% in identifying previously untrained in-app activity traffic as unknown data when this framework is employed.
Overlapping speech diarization has been traditionally treated as a multi-label classification problem. In this paper, we reformulate this task as a single-label prediction problem by encoding multiple binary labels into a single label with the power set, which represents the possible combinations of target speakers. This formulation has two benefits. First, the overlaps of target speakers are explicitly modeled. Second, threshold selection is no longer needed. Through this formulation, we propose the speaker embedding-aware neural diarization (SEND) framework, where a speech encoder, a speaker encoder, two similarity scorers, and a post-processing network are jointly optimized to predict the encoded labels according to the similarities between speech features and speaker embeddings. Experimental results show that SEND has a stable learning process and can be trained on highly overlapped data without extra initialization. More importantly, our method achieves the state-of-the-art performance in real meeting scenarios with fewer model parameters and lower computational complexity.