Telehealth and remote health monitoring have become increasingly important during the SARS-CoV-2 pandemic and it is widely expected that this will have a lasting impact on healthcare practices. These tools can help reduce the risk of exposing patients and medical staff to infection, make healthcare services more accessible, and allow providers to see more patients. However, objective measurement of vital signs is challenging without direct contact with a patient. We present a video-based and on-device optical cardiopulmonary vital sign measurement approach. It leverages a novel multi-task temporal shift convolutional attention network (MTTS-CAN) and enables real-time cardiovascular and respiratory measurements on mobile platforms. We evaluate our system on an Advanced RISC Machine (ARM) CPU and achieve state-of-the-art accuracy while running at over 150 frames per second which enables real-time applications. Systematic experimentation on large benchmark datasets reveals that our approach leads to substantial (20%-50%) reductions in error and generalizes well across datasets.

Fast and robust voice-activity detection is critical to efficiently process speech. While deep-learning based methods to detect voice have shown competitive accuracies, the best models in the literature incur over a 100 ms latency on commodity processors. Such delays are unacceptable for real-time speech processing. In this paper, we study the impact of lowering the representation precision of the neural-network weights and neurons on both the accuracy and delay of voice-activity detection. Based on a design-space exploration, we not only determine the optimal scaling strategy but also adjust the network structure to accommodate the new quantization levels. Through experiments conducted with real user data, we demonstrate that optimized deep neural networks with lower bit precisions outperform the state-of-the-art WebRTC voice-activity detector with 87x lower delay and 6.8% lower error rate.

Recent work has shown that fast, compact low-bitwidth neural networks can be surprisingly accurate. These networks use homogeneous binarization: all parameters in each layer or (more commonly) the whole model have the same low bitwidth (e.g., 2 bits). However, modern hardware allows efficient designs where each arithmetic instruction can have a custom bitwidth, motivating heterogeneous binarization, where every parameter in the network may have a different bitwidth. In this paper, we show that it is feasible and useful to select bitwidths at the parameter granularity during training. For instance a heterogeneously quantized version of modern networks such as AlexNet and MobileNet, with the right mix of 1-, 2- and 3-bit parameters that average to just 1.4 bits can equal the accuracy of homogeneous 2-bit versions of these networks. Further, we provide analyses to show that the heterogeneously binarized systems yield FPGA- and ASIC-based implementations that are correspondingly more efficient in both circuit area and energy efficiency than their homogeneous counterparts.

Cost and accessibility have impeded the adoption of spirometers (devices that measure lung function) outside clinical settings, especially in low-resource environments. Prior work, called SpiroSmart, used a smartphone's built-in microphone as a spirometer. However, individuals in low- or middle-income countries do not typically have access to the latest smartphones. In this paper, we investigate how spirometry can be performed from any phone-using the standard telephony voice channel to transmit the sound of the spirometry effort. We also investigate how using a 3D printed vortex whistle can affect the accuracy of common spirometry measures and mitigate usability challenges. Our system, coined SpiroCall, was evaluated with 50 participants against two gold standard medical spirometers. We conclude that SpiroCall has an acceptable mean error with or without a whistle for performing spirometry, and advantages of each are discussed.

We describe techniques that allow inexpensive, ultra-thin, battery-free Radio Frequency Identification (RFID) tags to be turned into simple paper input devices. We use sensing and signal processing techniques that determine how a tag is being manipulated by the user via an RFID reader and show how tags may be enhanced with a simple set of conductive traces that can be printed on paper, stencil-traced, or even hand-drawn. These traces modify the behavior of contiguous tags to serve as input devices. Our techniques provide the capability to use off-the-shelf RFID tags to sense touch, cover, overlap of tags by conductive or dielectric (insulating) materials, and tag movement trajectories. Paper prototypes can be made functional in seconds. Due to the rapid deployability and low cost of the tags used, we can create a new class of interactive paper devices that are drawn on demand for simple tasks. These capabilities allow new interactive possibilities for pop-up books and other papercraft objects.

We explore the design space of energy-neutral situated displays, which give physical presence to digital information. We investigate three central dimensions: energy sources, display technologies, and wireless communications. Based on the power implications from our analysis, we present a thin, wireless, photovoltaic-powered display that is quick and easy to deploy and capable of indefinite operation in indoor lighting conditions. The display uses a low-resolution e-paper architecture, which is 35 times more energy-efficient than smaller-sized high-resolution displays. We present a detailed analysis on power consumption, photovoltaic energy harvesting performance, and a detailed comparison to other display-driving architectures. Depending on the ambient lighting, the display can trigger an update every 1 -- 25 minutes and communicate to a PC or smartphone via Bluetooth Low-Energy.

This paper presents a new gesture-based human interface for natural robot control. Detailed activity of the user's hand and arm is acquired via a novel device, called the BioSleeve, which packages dry-contact surface electromyography (EMG) and an inertial measurement unit (IMU) into a sleeve worn on the forearm. The BioSleeve's accompanying algorithms can reliably decode as many as sixteen discrete hand gestures and estimate the continuous orientation of the forearm. These gestures and positions are mapped to robot commands that, to varying degrees, integrate with the robot's perception of its environment and its ability to complete tasks autonomously. This flexible approach enables, for example, supervisory point-to-goal commands, virtual joystick for guarded teleoperation, and high degree of freedom mimicked manipulation, all from a single device. The BioSleeve is meant for portable field use; unlike other gesture recognition systems, use of the BioSleeve for robot control is invariant to lighting conditions, occlusions, and the human-robot spatial relationship and does not encumber the user's hands. The BioSleeve control approach has been implemented on three robot types, and we present proof-of-principle demonstrations with mobile ground robots, manipulation robots, and prosthetic hands.