Cognitive workload affects operators' performance principally in high-risk or time-demanding situations and when multitasking is required. An online cognitive workload monitoring system can provide valuable inputs to decision-making instances, such as the operator's state of mind and resulting performance. Therefore, it can allow potential adaptive support to the operator. This work presents a new design of a wearable embedded system for online cognitive workload monitoring. This new wearable system consists of, on the hardware side, a multi-channel physiological signals acquisition (respiration cycles, heart rate, skin temperature, and pulse waveform) and a low-power processing platform. Further, on the software side, our wearable embedded system includes a novel energy-aware bio-signal processing algorithm. We also use the concept of application self-awareness to enable energy-scalable embedded machine learning algorithms and methods for online subjects' cognitive workload monitoring. Our results show that this new wearable system can continuously monitor multiple bio-signals, compute their key features, and provide reliable detection of high and low cognitive workload levels with a time resolution of 1 minute and a battery lifetime of 14.58h in our experimental conditions. It achieves a detection accuracy of 76.6% (2.6% lower than analogous offline computer-based analysis) with a sensitivity of 77.04% and a specificity of 81.75%, on a simulated drone rescue mission task. Moreover, by applying our self-aware monitoring to exploit different energy-scalable modes, we can increase battery lifetime by 51.6% (up to 22.11 hours) while incurring an insignificant accuracy loss of 1.07%.

No abstract

Cyber-Physical Systems (CPSs) are systems designed as a network of different interacting elements, which integrate computational and physical capabilities. The human-machine interaction plays a significant role in CPSs, especially in applications where people are an active element. In this context, emotion recognition is a relevant aspect to achieve a more efficient, collaborative, and resilient machine performance in collaboration with humans. On this basis, this paper proposes an embedded machine learning approach for emotion recognition fully implemented in an ultra low-power System-on-Chip (SoC) with limited resources. To this end, the intelligence system considers a reduced set of raw physiological signals within an approximate computing focus.

In the last years, the Industry 4.0 paradigm is gaining relevance in the agro-food industry, leading to Smart Farming. One of the applications in the Smart Farming domain is the advanced chemical analysis in process monitoring using distributed, low-cost embedded systems. Optical sensing technology is used in conjunction with machine learning techniques for this advanced analysis. From the embedded system perspective, it might be required to propose a method for the implementation of machine learning techniques in heterogeneous platforms. This paper focuses on implementing Machine Learning techniques in a System on Programmable Chip, based on an FPGA and ARM processors. As a use case, we mimic water pollution by ethanol. Thus, the application might determine the percentage of ethanol of the water during run-time. As a result, this paper provides a methodology for implementing a machine learning technique for ethanol prediction using an FPGA, and the study of its parameters as resource utilization and accelerator latency for the architecture proposed.

Nowadays, the growth of Industry 4.0 and Internet of Things (IoT) demands new solutions for designing low-power low-cost advanced computational algorithms. This work develops the sensor signal processing layer of a chemical biosensing IoT edge device using NanoPillar transducers. We propose to move from smart sensors to expert sensors, applying Principal Component Analysis (PCA) for dimensionality reduction in FPGAs. As a result, this paper provides a design space exploration of PCA implementation over FPGAs, studying parameters as throughput and resource usage.