Hardware Acceleration of EEG-Based Emotion Classification Systems: A Comprehensive Survey.

Recent years have witnessed a growing interest in EEG-based wearable classifiers of emotions, which could enable the real-time monitoring of patients suffering from neurological disorders such as Amyotrophic Lateral Sclerosis (ALS), Autism Spectrum Disorder (ASD), or Alzheimer's. The hope is that such wearable emotion classifiers would facilitate the patients' social integration and lead to improved healthcare outcomes for them and their loved ones. Yet in spite of their direct relevance to neuro-medicine, the hardware platforms for emotion classification have yet to fill up some important gaps in their various approaches to emotion classification in a healthcare context. In this paper, we present the first hardware-focused critical review of EEG-based wearable classifiers of emotions and survey their implementation perspectives, their algorithmic foundations, and their feature extraction methodologies. We further provide a neuroscience-based analysis of current hardware accelerators of emotion classifiers and use it to map out several research opportunities, including multi-modal hardware platforms, accelerators with tightly-coupled cores operating robustly in the near/supra-threshold region, and pre-processing libraries for universal EEG-based datasets.

Shoe-Integrated, Force Sensor Design for Continuous Body Weight Monitoring.

Traditional pedobarography methods use direct force sensor placement in the shoe insole to record pressure patterns. One problem with such methods is that they tap only a few points on the flat sole under the foot and, therefore, do not account for the total ground reaction force. As a result, body weight tends to be under-estimated. This disadvantage has made it more difficult for pedobarography to be used to monitor many diseases, especially when their symptoms include body weight changes. In this paper, the problem of pedobarographic body weight measurement is addressed using a novel ergonomic shoe-integrated sensor array architecture based on concentrating the applied force via three-layered structures that we call Sandwiched Sensor Force Consolidators (SSFC). A shoe prototype is designed with the proposed sensors and shown to accurately measure body weight with an achievable relative accuracy greater than 99%, even in the presence of motion. The achieved relative accuracy is at least 4X better than the existing state of the art. The SSFC shoe prototype is built using readily available soccer shoes and piezoresistive FlexiForce sensors. To improve the wearability and comfort of the instrumented shoe, a semi-computational sensor design methodology is developed based on an equivalent-area concept that can accurately account for SSFC's with arbitrary shapes. The search space of the optimal SSFC design is shown to be combinatorial, and a high-performance computing (HPC) framework based on OpenMP parallel programming is proposed to accelerate the design optimization process. An optimal sensor design speedup of up to 22X is shown to be achievable using the HPC implementation.

Editorial for the Special Issue on MEMS Accelerometers.

Micro-Electro-Mechanical Systems (MEMS) devices are widely used for motion, pressure, light, and ultrasound sensing applications [...].

EEG-based Emotion Detection Using Unsupervised Transfer Learning.

Emotion classification using EEG signal processing has the potential of significantly improving the social integration of patients suffering from neurological disorders such as Amyotrophic Lateral Sclerosis (ALS) or the acute stages of Alzheimer's disease. One important challenge to the implementation of high-fidelity emotion recognition systems is the inadequacy of EEG data in terms of Signal-to-noise ratio (SNR), duration, and subject-to-subject variability. In this paper, we present a novel, integrated framework for semi-generic emotion detection using (1) independent component analysis for EEG preprocessing, (2) EEG subject clustering by unsupervised learning, and (3) a convolutional neural network (CNN) for EEG-based emotion recognition. The training and testing data was built using the combination of two publicly available repositories (DEAP and DREAMER), and a local dataset collected at Khalifa University using the standard International Affective Picture System (IAPS). The CNN classifier with the proposed transfer learning approach achieves an average accuracy of 70.26% for valence and 72.42% for arousal, which are superior to the reported accuracies of all generic (subject-independent) emotion classifiers.

A Self-Synchronizing, Low-Power, Low-Complexity Transceiver for Body-Coupled Communication.

This paper presents a self-synchronizing, low-power, low-complexity body-coupled communication (BCC) transceiver using the recently proposed Pulsed-Index Communication (PIC) techniques. The unique features of these techniques are used to simplify the BCC transceiver hardware and reduce its power consumption by eliminating the need for circuitries dedicated to clock and data recovery (CDR) and duty cycle correction. The self-synchronizing feature of the transceiver is achieved by exploiting the edge-coding property of PIC which consists of using pulse edges for encoding and detecting transmitted pulses rather than bit times or duty cycles. A working prototype of the proposed BCC transceiver using off-the-shelf components is developed and used to test, for the first time, a full, bi-directional BCC link by transmitting arbitrary 16-bit data words through the human body over a range of 150cm with zero bit-error rate and sub-1nJ/bit energy efficiency.

Monolithic Multi Degree of Freedom (MDoF) Capacitive MEMS Accelerometers.

With the continuous advancements in microelectromechanical systems (MEMS) fabrication technology, inertial sensors like accelerometers and gyroscopes can be designed and manufactured with smaller footprint and lower power consumption. In the literature, there are several reported accelerometer designs based on MEMS technology and utilizing various transductions like capacitive, piezoelectric, optical, thermal, among several others. In particular, capacitive accelerometers are the most popular and highly researched due to several advantages like high sensitivity, low noise, low temperature sensitivity, linearity, and small footprint. Accelerometers can be designed to sense acceleration in all the three directions (X, Y, and Z-axis). Single-axis accelerometers are the most common and are often integrated orthogonally and combined as multiple-degree-of-freedom (MDoF) packages for sensing acceleration in the three directions. This type of MDoF increases the overall device footprint and cost. It also causes calibration errors and may require expensive compensations. Another type of MDoF accelerometers is based on monolithic integration and is proving to be effective in solving the footprint and calibration problems. There are mainly two classes of such monolithic MDoF accelerometers, depending on the number of proof masses used. The first class uses multiple proof masses with the main advantage being zero calibration issues. The second class uses a single proof mass, which results in compact device with a reduced noise floor. The latter class, however, suffers from high cross-axis sensitivity. It also requires very innovative layout designs, owing to the complicated mechanical structures and electrical contact placement. The performance complications due to nonlinearity, post fabrication process, and readout electronics affects both classes of accelerometers. In order to effectively compare them, we have used metrics such as sensitivity per unit area and noise-area product. This paper is devoted to an in-depth review of monolithic multi-axis capacitive MEMS accelerometers, including a detailed analysis of recent advancements aimed at solving their problems such as size, noise floor, cross-axis sensitivity, and process aware modeling.

A preliminary evaluation of continuous, shoe-integrated weight measurements for heart failure patients.

Weight measurements are part of current medical care for congestive heart failure (CHF) patients. In this work, we explore the potential of shoe-mounted pressure sensors to automatically and remotely estimate the weight of CHF patients. We show that weight estimation accuracy degrades due to human subject movement. Moreover, we show that for a standing human subject the accuracy is influenced by the characteristic of the sensor used to measure pressure. Our experimental evaluation of various pressure sensors widely used in wearable applications reveals that they exhibit properties that are undesirable for precise weight measurements.