Reducing the Energy Consumption of sEMG-Based Gesture Recognition at the Edge Using Transformers and Dynamic Inference.

Hand gesture recognition applications based on surface electromiographic (sEMG) signals can benefit from on-device execution to achieve faster and more predictable response times and higher energy efficiency. However, deploying state-of-the-art deep learning (DL) models for this task on memory-constrained and battery-operated edge devices, such as wearables, requires a careful optimization process, both at design time, with an appropriate tuning of the DL models' architectures, and at execution time, where the execution of large and computationally complex models should be avoided unless strictly needed. In this work, we pursue both optimization targets, proposing a novel gesture recognition system that improves upon the state-of-the-art models both in terms of accuracy and efficiency. At the level of DL model architecture, we apply for the first time tiny transformer models (which we call bioformers) to sEMG-based gesture recognition. Through an extensive architecture exploration, we show that our most accurate bioformer achieves a higher classification accuracy on the popular Non-Invasive Adaptive hand Prosthetics Database 6 (Ninapro DB6) dataset compared to the state-of-the-art convolutional neural network (CNN) TEMPONet (+3.1%). When deployed on the RISC-V-based low-power system-on-chip (SoC) GAP8, bioformers that outperform TEMPONet in accuracy consume 7.8×-44.5× less energy per inference. At runtime, we propose a three-level dynamic inference approach that combines a shallow classifier, i.e., a random forest (RF) implementing a simple "rest detector" with two bioformers of different accuracy and complexity, which are sequentially applied to each new input, stopping the classification early for "easy" data. With this mechanism, we obtain a flexible inference system, capable of working in many different operating points in terms of accuracy and average energy consumption. On GAP8, we obtain a further 1.03×-1.35× energy reduction compared to static bioformers at iso-accuracy.

Q-PPG: Energy-Efficient PPG-Based Heart Rate Monitoring on Wearable Devices.

Hearth Rate (HR) monitoring is increasingly performed in wrist-worn devices using low-cost photoplethysmography (PPG) sensors. However, Motion Artifacts (MAs) caused by movements of the subject's arm affect the performance of PPG-based HR tracking. This is typically addressed coupling the PPG signal with acceleration measurements from an inertial sensor. Unfortunately, most standard approaches of this kind rely on hand-tuned parameters, which impair their generalization capabilities and their applicability to real data in the field. In contrast, methods based on deep learning, despite their better generalization, are considered to be too complex to deploy on wearable devices. In this work, we tackle these limitations, proposing a design space exploration methodology to automatically generate a rich family of deep Temporal Convolutional Networks (TCNs) for HR monitoring, all derived from a single "seed" model. Our flow involves a cascade of two Neural Architecture Search (NAS) tools and a hardware-friendly quantizer, whose combination yields both highly accurate and extremely lightweight models. When tested on the PPG-Dalia dataset, our most accurate model sets a new state-of-the-art in Mean Absolute Error. Furthermore, we deploy our TCNs on an embedded platform featuring a STM32WB55 microcontroller, demonstrating their suitability for real-time execution. Our most accurate quantized network achieves 4.41 Beats Per Minute (BPM) of Mean Absolute Error (MAE), with an energy consumption of 47.65 mJ and a memory footprint of 412 kB. At the same time, the smallest network that obtains a MAE 8 BPM, among those generated by our flow, has a memory footprint of 1.9 kB and consumes just 1.79 mJ per inference.

On the impact of smart sensor approximations on the accuracy of machine learning tasks.

Smart sensors present in ubiquitous Internet of Things (IoT) devices often obtain high energy efficiency by carefully tuning how the sensing, the analog to digital (A/D) conversion and the digital serial transmission are implemented. Such tuning involves approximations, i.e. alterations of the sensed signals that can positively affect energy consumption in various ways. However, for many IoT applications, approximations may have an impact on the quality of the produced output, for example on the classification accuracy of a Machine Learning (ML) model. While the impact of approximations on ML algorithms is widely studied, previous works have focused mostly on processing approximations. In this work, in contrast, we analyze how the signal alterations imposed by smart sensors impact the accuracy of ML classifiers. We focus in particular on data alterations introduced in the serial transmission from a smart sensor to a processor, although our considerations can also be extended to other sources of approximation, such as A/D conversion. Results on several types of models and on two different datasets show that ML algorithms are quite resilient to the alterations produced by smart sensors, and that the serial transmission energy can be reduced by up to 70% without a significant impact on classification accuracy. Moreover, we also show that, contrarily to expectations, the two generic approximation families identified in our work yield similar accuracy losses.

LAPSE: Low-Overhead Adaptive Power Saving and Contrast Enhancement for OLEDs.

Organic Light Emitting Diode (OLED) display panels are becoming increasingly popular especially in mobile devices; one of the key characteristics of these panels is that their power consumption strongly depends on the displayed image. In this paper we propose LAPSE, a new methodology to concurrently reduce the energy consumed by an OLED display and enhance the contrast of the displayed image, that relies on image-specific pixel-by-pixel transformations. Unlike previous approaches, LAPSE focuses specifically on reducing the overheads required to implement the transformation at runtime. To this end, we propose a transformation that can be executed in real time, either in software, with low time overhead, or in a hardware accelerator with a small area and low energy budget. Despite the significant reduction in complexity, we obtain comparable results to those achieved with more complex approaches in terms of power saving and image quality. Moreover, our method allows to easily explore the full quality-versus-power tradeoff by acting on a few basic parameters; thus, it enables the runtime selection among multiple display quality settings, according to the status of the system.

The Human Brain Project and neuromorphic computing.

Understanding how the brain manages billions of processing units connected via kilometers of fibers and trillions of synapses, while consuming a few tens of Watts could provide the key to a completely new category of hardware (neuromorphic computing systems). In order to achieve this, a paradigm shift for computing as a whole is needed, which will see it moving away from current "bit precise" computing models and towards new techniques that exploit the stochastic behavior of simple, reliable, very fast, lowpower computing devices embedded in intensely recursive architectures. In this paper we summarize how these objectives will be pursued in the Human Brain Project.