Adopting Graph Neural Networks to Analyze Human-Object Interactions for Inferring Activities of Daily Living.

Human Activity Recognition (HAR) refers to a field that aims to identify human activities by adopting multiple techniques. In this field, different applications, such as smart homes and assistive robots, are introduced to support individuals in their Activities of Daily Living (ADL) by analyzing data collected from various sensors. Apart from wearable sensors, the adoption of camera frames to analyze and classify ADL has emerged as a promising trend for achieving the identification and classification of ADL. To accomplish this, the existing approaches typically rely on object classification with pose estimation using the image frames collected from cameras. Given the existence of inherent correlations between human-object interactions and ADL, further efforts are often needed to leverage these correlations for more effective and well justified decisions. To this end, this work proposes a framework where Graph Neural Networks (GNN) are adopted to explicitly analyze human-object interactions for more effectively recognizing daily activities. By automatically encoding the correlations among various interactions detected through some collected relational data, the framework infers the existence of different activities alongside their corresponding environmental objects. As a case study, we use the Toyota Smart Home dataset to evaluate the proposed framework. Compared with conventional feed-forward neural networks, the results demonstrate significantly superior performance in identifying ADL, allowing for the classification of different daily activities with an accuracy of 0.88. Furthermore, the incorporation of encoded information from relational data enhances object-inference performance compared to the GNN without joint prediction, increasing accuracy from 0.71 to 0.77.

A Reconfigurable Near-Sensor Processor for Anomaly Detection in Limb Prostheses.

This paper presents a reconfigurable near-sensor anomaly detection processor to real-time monitor the potential anomalous behaviors of amputees with limb prostheses. The processor is low-power, low-latency, and suitable for equipment on the prostheses and comprises a reconfigurable Variational Autoencoder (VAE), a scalable Self-Organizing Map (SOM) Array, and a window-size-adjustable Markov Chain, which can implement an integrated miniaturized anomaly detection system. With the reconfigurable VAE, the proposed processor can support up to 64 sensor sampling channels programmable by global configuration, which can meet the anomaly detection requirements in different scenarios. A scalable SOM array allows for the selection of different sizes based on the complexity of the data. Unlike traditional time accumulation-based anomaly detection methods, the Markov Chain is utilized to detect time-series-based anomalous data. The processor is designed and fabricated in a UMC 40-nm LP technology with a core area of 1.49 mm2 and a power consumption of 1.81 mW. It achieves real-time detection performance with 0.933 average F1 Score for the FSP dataset within 24.22 µs, and 0.956 average F1 Score for the SFDLA-12 dataset within 30.48 µs, respectively. The energy dissipation of detection for each input feature is 43.84 nJ with the FSP dataset, and 55.17 nJ with the SFDLA-12 dataset. Compared with ARM Cortex-M4 and ARM Cortex-M33 microcontrollers, the processor achieves energy and area efficiency improvements ranging from 257×, 193× and 11×, 8×, respectively.

A Sensor-Based Decision Support System for Transfemoral Socket Rectification.

A decision support system (DSS) was developed that outputs suggestions for socket-rectification actions to the prosthetist, aiming at improving the fitness of transfemoral prosthetic socket design and reducing the time needed for the final socket design. For this purpose, the DSS employs a fuzzy-logic inference engine (IE) which combines a set of rectification rules with pressure measurements generated by sensors embedded in the socket, for deciding the rectification actions. The latter is then processed by an algorithm that receives, manipulates and modifies a 3D digital socket model as a triangle mesh formatted inside an STL file. The DSS results were validated and tested in an FEA simulation environment, by simulating and comparing the donning process among a good-fitting socket, a loose socket (poor-fit) and several rectified sockets produced by the proposed DSS. The simulation results indicate that volume reduction improves the pressure distribution over the stump. However, as the intensity of socket rectification increases, i.e., as volume reduction increases, high pressures appear in other parts of the socket which generate discomfort. Therefore, a trade-off is required between the amount of rectification and the balance of the pressure distributions experienced at the stump.