Getting a good feature representation of data is paramount for Human Activity Recognition (HAR) using wearable sensors. An increasing number of feature learning approaches—in particular deep-learning based—have been proposed to extract an effective feature representation by analyzing large amounts of data. However, getting an objective interpretation of their performances faces two problems: the lack of a baseline evaluation setup, which makes a strict comparison between them impossible, and the insufficiency of implementation details, which can hinder their use. In this paper, we attempt to address both issues: we firstly propose an evaluation framework allowing a rigorous comparison of features extracted by different methods, and use it to carry out extensive experiments with state-of-the-art feature learning approaches. We then provide all the codes and implementation details to make both the reproduction of the results reported in this paper and the re-use of our framework easier for other researchers. Our studies carried out on the OPPORTUNITY and UniMiB-SHAR datasets highlight the effectiveness of hybrid deep-learning architectures involving convolutional and Long-Short-Term-Memory (LSTM) to obtain features characterising both short- and long-term time dependencies in the data.

The scarcity of labelled time-series data can hinder a proper training of deep learning models. This is especially relevant for the growing field of ubiquitous computing, where data coming from wearable devices have to be analysed using pattern recognition techniques to provide meaningful applications. To address this problem, we propose a transfer learning method based on attributing sensor modality labels to a large amount of time-series data collected from various application fields. Using these data, our method firstly trains a Deep Neural Network (DNN) that can learn general characteristics of time-series data, then transfers it to another DNN designed to solve a specific target problem. In addition, we propose a general architecture that can adapt the transferred DNN regardless of the sensors used in the target field making our approach in particular suitable for multichannel data. We test our method for two ubiquitous computing problems—Human Activity Recognition (HAR) and Emotion Recognition (ER)—and compare it a baseline training the DNN without using transfer learning. For HAR, we also introduce a new dataset, Cognitive Village-MSBand (CogAge), which contains data for 61 atomic activities acquired from three wearable devices (smartphone, smartwatch, and smartglasses). Our results show that our transfer learning approach outperforms the baseline for both HAR and ER.

Identifying accident patterns is one of the most vital research foci of driving analysis. Environmental or safety applications and the growing area of fleet management all benefit from accident detection contributions by minimizing the risk vehicles and drivers are subject to, improving their service and reducing overhead costs. Some solutions have been proposed in the past literature for automated accident detection that are mainly based on traffic data or external sensors. However, traffic data can be difficult to access, while external sensors can end up being difficult to set up and unreliable, depending on how they are used. Additionally, the scarcity of accident detection data has limited the type of approaches used in the past, leaving in particular, machine learning (ML) relatively unexplored. Thus, in this paper, we propose a ML framework for automated car accident detection based on mutimodal in-car sensors. Our work is a unique and innovative study on detecting real-world driving accidents by applying state-of-the-art feature extraction methods using basic sensors in cars. In total, five different feature extraction approaches, including techniques based on feature engineering and feature learning with deep learning are evaluated on the strategic highway research program (SHRP2) naturalistic driving study (NDS) crash data set. The main observations of this study are as follows: (1) CNN features with a SVM classifier obtain very promising results, outperforming all other tested approaches. (2) Feature engineering and feature learning approaches were finding different best performing features. Therefore, our fusion experiment indicates that these two feature sets can be efficiently combined. (3) Unsupervised feature extraction remarkably achieves a notable performance score.

While even the most common definition of pain is under debate, pain assessment has remained the same for decades. But the paramount importance of precise pain management for successful healthcare has encouraged initiatives to improve the way pain is assessed. Recent approaches have proposed automatic pain evaluation systems using machine learning models trained with data coming from behavioural or physiological sensors. Although yielding promising results, machine learning studies for sensor-based pain recognition remain scattered and not necessarily easy to compare to each other. In particular, the important process of extracting features is usually optimised towards specific datasets. We thus introduce a comparison of feature extraction methods for pain recognition based on physiological sensors in this paper. In addition, the PainMonit Database (PMDB), a new dataset including both objective and subjective annotations for heat-induced pain in 52 subjects, is introduced. In total, five different approaches including techniques based on feature engineering and feature learning with deep learning are evaluated on the BioVid and PMDB datasets. Our studies highlight the following insights: (1) Simple feature engineering approaches can still compete with deep learning approaches in terms of performance. (2) More complex deep learning architectures do not yield better performance compared to simpler ones. (3) Subjective self-reports by subjects can be used instead of objective temperature-based annotations to build a robust pain recognition system.

Background: Flow experience is a specific positive and affective state that occurs when humans are completely absorbed in an activity and forget everything else. This state can lead to high performance, well-being, and productivity at work. Few studies have been conducted to determine the human flow experience using physiological wearable sensor devices. Other studies rely on self-reported data. Methods: In this article, we use physiological data collected from 25 subjects with multimodal sensing devices - the Empatica E4 wristband, the Emotiv Epoc X electroencephalography (EEG) headset, and the Biosignalplux RespiBAN - in arithmetic and reading tasks to automatically discriminate between flow and non-flow states using feature engineering and deep feature learning approaches. The most meaningful wearable device for flow detection is determined by comparing the performances of each device. We also investigate the connection between emotions and flow by testing transfer learning techniques involving an emotion recognition-related task on the source domain. Results: The EEG sensor modalities yielded the best performances with an accuracy of 64.97%, and a macro Averaged F1 (AF1) score of 64.95%. An accuracy of 73.63% and an AF1 score of 72.70% were obtained after fusing all sensor modalities from all devices. Additionally, our proposed transfer learning approach using emotional arousal classification on the DEAP dataset led to an increase in performances with an accuracy of 75.10% and an AF1 score of 74.92%. Conclusion: The results of this study suggest that effective discrimination between flow and non-flow states is possible with multimodal sensor data. The success of transfer learning using the DEAP emotion dataset as a source domain indicates that emotions and flow are connected, and emotion recognition can be used as a latent task to enhance the performance of flow recognition.