In this work, we present a cloud-based system for noncontact, real-time recognition, and monitoring of physical activities and walking periods within a domestic environment. The proposed system employs standalone Internet of Things (IoT)-based millimeter wave radar devices and deep learning models to enable autonomous, free-living activity recognition, and gait analysis. To train deep learning models, we utilize range-Doppler maps generated from a data set of real-life in-home activities. The performance of several deep learning models is evaluated based on accuracy and prediction time, with the gated recurrent network [gated recurrent unit (GRU)] model selected for real-time deployment due to its balance of speed and accuracy compared to 2-D convolutional neural network long short-term memory (2D-CNNLSTM) and long short-term memory (LSTM) models. The overall accuracy of the GRU model for classifying in-home physical activities of trained subjects is 93%, with 86% accuracy for a new subject. In addition to recognizing and differentiating various activities and walking periods, the system also records the subject’s activity level over time, washroom use frequency, sleep/sedentary/active/out-of-home durations, current state, and gait parameters. Importantly, the system maintains privacy by not requiring the subject to wear or carry any additional devices.

We propose a novel algorithm to identify occupied seats in a motor vehicle, i.e., the number of occupants and their positions, using a frequency modulated continuous wave radar. Instead of using a high-resolution radar, which increases the cost and device size, and performing complex signal processing with several variables to be tuned for each scenario, we integrate machine learning algorithms with a low-cost radar system. Based on heat maps obtained from the Capon beamformer, we train a machine classifier to predict the number of occupants and their positions in a vehicle. We follow two different classification methods: multiclass classification and binary classification. We compare three classifiers: support vector machine (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF), in terms of accuracy and computational complexity for both testing and training sets. Our proposed system using an SVM classifier achieved an overall accuracy of 97% in classifying the defined scenarios in both multiclass classification and binary classification methods. In addition, to show the effectiveness of our proposed in-vehicle occupancy detection method, we provide the results of a commonly available people counting and tracking method for occupancy detection. Compared to common methods, the effectiveness, robustness, and accuracy of our proposed in-vehicle occupancy detection method are demonstrated.

We propose a novel corridor or hallway gait monitoring system based on radar signal processing, unsupervised learning, and a subject detection, association and tracking method. This paper proposes an algorithm that could be paired with any type of MIMO FMCW radar to capture human gait in a highly cluttered environment without needing radar antenna alteration. We validate algorithm functionality by capturing spatiotemporal gait values (e.g., speed, step points, step time, step length, and step count) of people walking in a long hallway. We show that our proposed algorithm yields an average absolute error for speed estimation between 0.0040 m/s to 0.0435 m/s. These preliminary results demonstrate the promising potential of our algorithm to accurately monitor gait in hallways, which increases opportunities for its applications in institutional and home environments.

In this paper, we use a multiple-input multiple-output (MIMO) radar for in-vehicle passenger detection. We propose a 2D CNN-LSTM (Convolutional Neural Network- Long Short-Term Memory) to accurately detect, count and classify passengers inside five-seater vehicles. Our deep learning model first extracts feature using a CNN model, and then a series of frames will be delivered to a time series model (LSTM) to predict new scenarios. Additionally, we provide the outcomes of various deep learning models and show that temporal deep learning models perform better in our radar datasets. Furthermore, we provide reliable session-dependent datasets collected from different car models with various passengers (including infants/ children and adults). The results show that our proposed 2D CNN-LSTM model can detect unattended infants/ children in vehicles with more than 95% accuracy, count passengers and identify their occupied seats with an accuracy of 89% and classify passengers with more than 74% accuracy. Since our model is evaluated in a new car with new passengers, it ensures the generality of our proposed method to be deployed in any five-seater vehicle.

According to the group Kids and Cars, since 1990, nearly 1000 kids lost their lives because they were deliberately or unintentionally left in parked vehicles to potentially overheat or freeze. The development of technology able to prevent and address this serious, worldwide problem is crucial. In this paper, we deploy a radar‐based sensor for in‐vehicle presence‐absence detection of a living body. We present a novel radar signal processing technique to identify the presence or absence of a living body in a vehicle using a mm‐wave frequency‐modulated continuous‐wave (FMCW) radar. Our proposed method is based on reflections from breathing cycles creating correlated and consistent micro‐Doppler effects over time. The performance of the system is evaluated with adults and two phantoms mimicking the breathing of children in various scenarios. The results show that we can clearly detect any tiny living body in vehicles with 100% accuracy without a need for any compute‐intensive complex signal processing, making the system of extreme low‐cost. The results demonstrate the high sensitivity and robustness of the mm‐wave system in extensive studies over the course of multiple months.

In the present investigation, the impacts of antenna field of view (FOV) on the accuracy of machine learning (ML) models utilized for the classification of various unmanned air vehicle (UAV) types were systematically explored using full-wave electromagnetic simulation software. Initially, similar to many state-of-the-art works, an ML algorithm was meticulously trained under a particular condition where the relative angle between the UAVs and the antenna was kept at 0 $^{\circ}$ , yielding an accuracy of 96.5%. In contrast to the common practice, the trained ML model was subsequently subjected to testing at varied relative angles, spanning from 20 $^{\circ}$ to 90 $^{\circ}$ . Observational outcomes delineate a decrease in ML classification accuracy with an increase in the relative angle between UAVs and the antenna. Next, this study investigates the impact of utilizing multiple-input-multiple-output (MIMO) radar system for classification. The results indicate enhancement of UAV detection and classification efficacy when contrasted with a single-input-single-output (SISO) radar system, at 80 $^{\circ}$ , an accuracy of 60% for the MIMO antenna compared to 13% for the SISO antenna. To corroborate the results obtained by utilizing full-wave electromagnetic simulation software, a series of experimental laboratory measurements were undertaken. The empirical measurements were found to yield comparable ML accuracy, at 80 $^{\circ}$ , an accuracy of 38% for the MIMO antenna compared to 1% for the SISO antenna. This work underscores the paramount versatility achieved through invoking full-wave electromagnetic simulators alongside ML algorithms and the respective possible impact on standard practices when it comes to developing next-generation MIMO radar systems for UAV classification.