Search

Deep learning models have achieved significant success in human activity recognition, particularly in assisted living and telemonitoring. However, training these models requires substantial amounts of labeled training data, which is time-consuming and costly to acquire in real-world environments. Contrastive self-supervised learning has recently garnered attention in sensor-based activity recognition to mitigate the need for expensive large-scale data collection and annotation. Despite numerous related published papers, there remains a lack of literature reviews highlighting recent advances in contrastive self-supervised learning for sensor-based activity recognition. This paper extensively reviews 43 papers on recent contrastive self-supervised learning methods for sensor-based human activity recognition, excluding those related to video or audio sensors due to privacy concerns. First, we summarize the taxonomy of contrastive self-supervised learning, followed by a detailed description of contrastive learning models used for activity recognition and their main components. Next, we comprehensively review data augmentation methods for sensor data and commonly used benchmark datasets for activity recognition. The empirical performance comparisons of different methods are presented on benchmark datasets in linear evaluation, semi-supervised learning, and transfer learning scenarios. Through these comparisons, we derive significant insights into the selection of contrastive self-supervised models for sensor-based activity recognition. Finally, we discuss the limitations of current research and outline promising research directions for future exploration.

Human Activity Recognition (HAR) is one of the most popular research topics thanks to its usefulness in providing targeted, meaningful assistance to older adults. Because of the aging of the population in first-world countries, it becomes increasingly important to find innovative solutions that reduce risks associated with aging-in-place policies. HAR proposes solutions that are based on Ambient Intelligence (AmI) to alleviate those risks. In this work, we exploited three UWB radars to recognize 14 activities performed by 19 participants in a prototype smart-home apartment. The main contribution of this paper is UWB radar data cleaning on a practical dataset. The UWB radar data has been filtered using an unsupervised deep convolutional autoencoder (CNN-AE) that learns background noise from the data. This filtering method is compared to the unfiltered data using a Convolutional Neural Network (CNN) classifier in a Leave-One-Subject-Out (LOSO) classification. Performances attest that the CNN-AE unsupervised filtering is efficient for HAR. In addition, we tested the generalization potential of this architecture when the dataset is comprised of a lower number of participants (1, 5, 10, and all 19 participants). Generalization in HAR is difficult as the results show the importance of data quantity and number of subjects. We obtained 69.9% top-1 accuracy when using our filtering architecture compared to 48.4% without it. To conclude, we show that an unsupervised CNN-AE can efficiently filter and generalize UWB radar data in a HAR setting while providing easier learning constraints and implementation on a practical dataset.

Myotonic dystrophy type 1 (DM1) is a rare disease where the highest prevalence is found in the small geographical region of Saguenay-Lac-St-Jean in Quebec, Canada. This disease impacts the quality of life and the ability of the affected people to pursue their normal day to day activities. To attenuate the impact of DM1, one could suggest that physiotherapists or other health professionals prescribe adapted physical exercise programs to carry out at home. It is, however, unpractical for the professionals to monitor every patient during their program and prohibitive in terms of cost. One alternative solution is to rely on the use of ambient technologies. To do so, in this research, our team developed an assistive system able to recognize simple mobility related activities (MA) and monitor each exercise performed during training sessions. The system can help the person by providing guidance and motivation throughout the training. The system was tested on 10 persons affected by the disease in their home for 10 weeks. The results obtained are encouraging and we discuss them in comparison with our previous work conducted in lab settings. Finally, to help further advances the research, the datasets are openly available online to the community.

Polysomnography (PSG) is the standard test for diagnosing sleep apnea. However, the approach is obtrusive, time-consuming, and with limited access for patients in need of sleep apnea diagnosis. In recent years, there have been many attempts to search for an alternative device or approach that avoids the limitations of PSG. Pressure-sensitive mats (PSM) have proven to be able to detect central sleep apneas (CSA) and be a potential alternative for PSG. In the current study, we combine advanced machine learning approaches with a practical unobtrusive home monitoring device (PSM) to detect CSA events from data collected nocturnally and unattended. Two deep learning methods are implemented for the automatic detection of CSA events: a temporal convolutional network (TCN) and a bidirectional long short-term memory (BiLSTM) network. The deep learning models are compared to a classical machine learning approach (linear support vector machine, SVM) and a simple threshold-based algorithm. Considering the characteristics of each method, we choose strategies, including resampling and weighted cost-functions, to optimize the methods and to perform CSA detection as anomaly detection in an imbalanced data set. We evaluate the performance of all models on a database containing 7 days of data from 9 elderly patients. From the resulting 63 days, data from 7 patients (49 days) are devoted to training for optimizing hyperparameters, and data from 2 patients (14 days) are devoted to testing. Experimental results indicate that the best-performing model achieves an accuracy of 95.1% through training an BiLSTM network. Overall, the implemented deep learning methods achieve better performance than the conventional classification approach (SVM) and the simple threshold-based method, and show good potential for the use of PSM for practical unobtrusive monitoring of CSA.

Most supervised audio recognition systems developed to this point have used a testing set which includes the same categories as the training set database. Such systems are called closed-set recognition (CSR). However, audio recognition in real applications can be more complicated, where the datasets can be dynamic, and novel categories can ceaselessly be detected. Hence, in practice, the usual methods will assign to these novel classes labels which are often incorrect. This work aims to investigate audio open-set recognition (OSR) suitable for multi-classes classification recognition, with a rejection option for classes never seen by the system. A probabilistic calibration of a support vector machine classifier is utilized and formulated under the open-set scenario. For this, it is proposed to apply a threshold technique called peak side ratio (PSR) to the audio recognition task. A candidate label is first examined by a Platt-calibrated support vector machine (SVM) to produce posterior probabilities. The PSR is then used to characterize the distribution of posterior probabilities values. This process helps to determine a threshold in order to reject or accept a particular class. Our proposed method is evaluated on different variations of open sets, using well-known metrics. Experimental results reveal that our proposed method outperforms previous OSR approaches over a wide range of openness values.

This work introduces new binaural beamforming algorithms for hearing aids, with a good robustness to errors in the estimated target source propagation model. Two different binaural beamforming designs are proposed. One design is based on an adaptive null positioning scheme, and the second is based on a combination of the adaptive null positioning scheme and wider beampatterns. Simulations are performed using signals and propagation models obtained from multichannel binaural hearing aids recordings, including some in a mildly reverberant environment. Evaluations are done in terms of noise reduction and target distortion. Evaluation results illustrate the robustness of the two proposed designs to errors between the true and estimated directions of arrival for the target source, and to mismatch between the anechoic propagation models used for the beamformers designs and the reverberant propagation models used to generate the signals at the sensors or beamformer inputs. Both designs surpass the performance of standard binaural Minimum Variance Distortionless Response (MVDR) and binaural Generalized-Side Lobe Canceler (GSC) beamformers.