Precise microfluidic control is essential for biomedical diagnostics. Compared to traditional PCB-based DMFB, standard CMOS-based digital microfluidic biochips (DMFB) can integrate sensing, heating, and actuation on a single chip, eliminating the need for external equipment [1]. However, CMOS-DMFB are limited by their lower actuation voltage (usually around 90 V [2]), and their actuation stability is often not as good as that of PCB-DMFB, which can be operated at voltages as high as 300 V [3]. This is due to the fact that, according to research, the EWOD force is proportional to the square of the actuation voltage [4]. In other words, a CMOS-DMFB will only generate approximately 4% of the actuation force of a PCB-DMFB, making actuation not as easy as expected [5]. To alleviate this limitation, we would like to utilize the real-time sensing function on the chip as an actuation feedback. After each actuation, the sensing results can be used to re-adjust the path and determine whether the destination has been reached. Therefore, we designed M* algorithms for on-chip microfluidic operation using A* algorithms to improve the actuation efficiency and accuracy, which is important for high-precision biomedical applications.

The demand of high-resolution video contents has grown over the years. However, the delivery of high-resolution video is constrained by either computational resources required for rendering or network bandwidth for remote transmission. To remedy this limitation, we leverage the eye trackers found alongside existing augmented and virtual reality headsets. We propose the application of video super-resolution (VSR) technique to fuse low-resolution context with regional high-resolution context for resource-constrained consumption of high-resolution content without perceivable drop in quality. Eye trackers provide us the gaze direction of a user, aiding us in the extraction of the regional high-resolution context. As only pixels that falls within the gaze region can be resolved by the human eye, a large amount of the delivered content is redundant as we can’t perceive the difference in quality of the region beyond the observed region. To generate a visually pleasing frame from the fusion of high-resolution region and low-resolution region, we study the capability of a deep neural network of transferring the context of the observed region to other regions (low-resolution) of the current and future frames. We label this task a Foveated Video Super-Resolution (FVSR), as we need to super-resolve the low-resolution regions of current and future frames through the fusion of pixels from the gaze region. We propose Cross-Resolution Flow Propagation (CRFP) for FVSR. We train and evaluate CRFP on REDS dataset on the task of 8× FVSR, i.e. a combination of 8× VSR and the fusion of foveated region. Departing from the conventional evaluation of per frame quality using SSIM or PSNR, we propose the evaluation of past foveated region, measuring the capability of a model to leverage the noise present in eye trackers during FVSR. Code is made available at https://github.com/eugenelet/CRFP.

Deep neural networks (DNNs) are known to perform well when deployed to test distributions that shares high similarity with the training distribution. Feeding DNNs with new data sequentially that were unseen in the training distribution has two major challenges — fast adaptation to new tasks and catastrophic forgetting of old tasks. Such difficulties paved way for the on-going research on few-shot learning and continual learning. To tackle these problems, we introduce Attentive Independent Mechanisms (AIM). We incorporate the idea of learning using fast and slow weights in conjunction with the decoupling of the feature extraction and higher-order conceptual learning of a DNN. AIM is designed for higher-order conceptual learning, modeled by a mixture of experts that compete to learn independent concepts to solve a new task. AIM is a modular component that can be inserted into existing deep learning frameworks. We demonstrate its capability for few-shot learning by adding it to SIB and trained on MiniImageNet and CIFAR-FS, showing significant improvement. AIM is also applied to ANML and OML trained on Omniglot, CIFAR-100 and MiniImageNet to demonstrate its capability in continual learning. Code made publicly available at https://github.com/huang50213/AIM-Fewshot-Continual.

Deciding the amount of neurons during the design of a deep neural network to maximize performance is not intuitive. In this work, we attempt to search for the neuron (filter) configuration of a fixed network architecture that maximizes accuracy. Using iterative pruning methods as a proxy, we parametrize the change of the neuron (filter) number of each layer with respect to the change in parameters, allowing us to efficiently scale an architecture across arbitrary sizes. We also introduce architecture descent which iteratively refines the parametrized function used for model scaling. The combination of both proposed methods is coined as NeuralScale. To prove the efficiency of NeuralScale in terms of parameters, we show empirical simulations on VGG11, MobileNetV2 and ResNet18 using CIFAR10, CIFAR100 and TinyImageNet as benchmark datasets. Our results show an increase in accuracy of 3.04%, 8.56% and 3.41% for VGG11, MobileNetV2 and ResNet18 on CIFAR10, CIFAR100 and TinyImageNet respectively under a parameter-constrained setting (output neurons (filters) of default configuration with scaling factor of 0.25).

The adoption of biomedical signals such as photoplethysmogram (PPG) and electrocardiogram (ECG) for health parameter estimation on wearable devices is growing in tandem with the increase of attention in mobile healthcare. In our work, we use PPG signals extracted from PPG sensors which are used for biometric identification. A challenge for biometric identification using PPG signal is the variation in domain (placement of sensors, wavelengths, device variation, etc.). In this work, we propose the use of both unsupervised and semi-supervised adversarial learning techniques for cross-domain adaptation. As such algorithm will be deployed on wearable devices, we propose a compact model meeting tight memory footprint limitation. All experiments will be simulated using a public dataset (TROIKA) and our in-house dataset. By introducing a cross-domain adaptation approach across sensors, we observe an accuracy gain of 4.15% on our in-house dataset. The proposed semi-supervised learning technique gives an additional accuracy boost of 2.02%.

Signal acquisition from the wrist using wearable device is usually corrupted by noise, sometimes up to a level where the noise completely dominates the signal. In this paper, we use the acquisition of photoplethysmography (PPG) signal as an example to demonstrate our proposed algorithm implemented on a sensor array. To increase the chances of acquiring clean PPG signal, we develop a uniform linear sensor array located above the radial artery. We also develop a wearable device to integrate our sensor array and our proposed algorithm. We propose centralized state sensing (CSS), a centralized algorithm suited to our sensor array, to increase the efficiency in multi-sensor estimation. To support our proposed algorithm, we use heart rate estimation as an example and compare it with the estimation of a single sensor and the statistical mean of the estimation of multiple sensors. Experimental results demonstrate that our proposal achieve a reduction in mean absolute error of 26.8%, compared to the estimation result obtained using the average of the estimation of all sensors. Operating on a 80 MHz processor, our algorithm introduces a 8 ms overhead, making it highly suitable for wearable devices.