Rafael Luiz da Silva

Body-rocking is an undesired stereotypical motor movement performed by some individuals, and its detection is essential for self-awareness and habit change. We envision a pipeline that includes inertial wearable sensors and a real-time detection system for notifying the user so that they are aware of their body-rocking behavior. For this task, similarities of body rocking to other non-related repetitive activities may cause false detections which prevent continuous engagement, leading to alarm fatigue. We present a pipeline using Bayesian Neural Networks with uncertainty quantification for jointly reducing false positives and providing accurate detection. We show that increasing model capacity does not consistently yield higher performance by itself, while pairing it with the Bayesian approach does yield significant improvements. Disparities in uncertainty quantification are better quantified by calibrating them using deep neural networks. We show that the calibrated probabilities are effective quality indicators of reliable predictions. Altogether, we show that our approach provides additional insights on the role of Bayesian techniques in deep learning as well as aids in accurate body-rocking detection, improving our prior work on this subject.

Preprint of manuscript submitted to an IEEE Journal currently under revision.

Environmental context prediction is important for wearable robotic applications, such as terrain-adaptive control. System efficiency is critical for wearable robots, in which system resources (e.g., processors and memory) are highly constrained. This article aims to address the system efficiency of real-time environmental context prediction for lower limb prostheses. First, we develop an uncertainty-aware frame selection strategy that can dynamically select frames according to lower limb motion and uncertainty captured by Bayesian neural networks (BNNs) for environment prediction. We further propose a dynamic Bayesian gated recurrent unit (D-BGRU) network to address the inconsistent frame rate which is a side effect of the dynamic frame selection. Second, we investigate the effects on the tradeoff between computational complexity and environment prediction accuracy of adding additional sensing modalities (e.g., GPS and an on-glasses camera) into the system. Finally, we implement and optimize our framework for embedded hardware, and evaluate the real-time inference accuracy and efficiency of classifying six types of terrains. The experiments show that our proposed frame selection strategy can reduce more than 90% of the computations without sacrificing environment prediction accuracy, and can be easily extended to the situation of multimodality fusion. We achieve around 93% prediction accuracy with less than one frame to be processed per second. Our model has 6.4 million 16-bit float numbers and takes 44 ms to process each frame on a lightweight embedded platform (NVIDIA Jetson TX2).

The performance of a low-power single-lead armband in generating electrocardiogram (ECG) signals from the chest and left arm was validated against a BIOPAC MP160 benchtop system in real-time. The filtering performance of three adaptive filtering algorithms, namely least mean squares (LMS), recursive least squares (RLS), and extended kernel RLS (EKRLS) in removing white (W), power line interference (PLI), electrode movement (EM), muscle artifact (MA), and baseline wandering (BLW) noises from the chest and left-arm ECG was evaluated with respect to the mean squared error (MSE). Filter parameters of the used algorithms were adjusted to ensure optimal filtering performance. LMS was found to be the most effective adaptive filtering algorithm in removing all noises with minimum MSE. However, for removing PLI with a maximal signal-to-noise ratio (SNR), RLS showed lower MSE values than LMS when the step size was set to 1 × 10⁻⁵. We proposed a transformation framework to convert the denoised left-arm and chest ECG signals to their low-MSE and high-SNR surrogate chest signals. With wide applications in wearable technologies, the proposed pipeline was found to be capable of establishing a baseline for comparing left-arm signals with original chest signals, getting one step closer to making use of the left-arm ECG in clinical cardiac evaluations.

Reliable environmental context prediction is critical for wearable robots (e.g., prostheses and exoskeletons) to assist terrain-adaptive locomotion. This article proposed a novel vision-based context prediction framework for lower limb prostheses to simultaneously predict human's environmental context for multiple forecast windows. By leveraging the Bayesian neural networks (BNNs), our framework can quantify the uncertainty caused by different factors (e.g., observation noise, and insufficient or biased training) and produce a calibrated predicted probability for online decision-making. We compared two wearable camera locations (a pair of glasses and a lower limb device), independently and conjointly. We utilized the calibrated predicted probability for online decision-making and fusion. We demonstrated how to interpret deep neural networks with uncertainty measures and how to improve the algorithms based on the uncertainty analysis. The inference time of our framework on a portable embedded system was less than 80 ms/frame. The results in this study may lead to novel context recognition strategies in reliable decision-making, efficient sensor fusion, and improved intelligent system design in various applications.

We combined adaptive filters associated with power estimators to characterize the knock signal, obtained from a knock sensor, in an internal combustion engine. The filters were implemented using the automotive model-based design methodology, and the resulting software was embedded into hardware for a real-time evaluation of the proposed solution. The knock signals could be qualitatively identified in real-time, and thus have the potential to aid in the management of flexible-fuel engines. This approach has an extensive range of applications within the automotive industry, since it can be implemented within any model-based control strategy. For example, this methodology can be applied in commercial ECUs, currently used in most vehicles for knock detection, by simply eliminating an internal, dedicated, integrated circuit for knock identification, or by serving as a redundancy device (i.e: for safety purposes). Finally, this methodology can identify the knock signal, obtained from a knock sensor, just by using an algorithm implemented in the ECU’s processor, which we show is identifiable in real-time.