30 October 2023 in IEEE Transactions on Circuits and Systems II: Express Briefs
Radar-based human activity recognition (HAR) has great potential in many applications. The deployment of HAR in many practical applications requires the system to be energy-resource efficient. In this work, we propose a lightweight dedicated hardware accelerator for radar-based HAR. In particular, the specific property of radar signals are taken advantages of to optimize the 1-D convolutional neural network (CNN) model for HAR. Based on that, a dedicated hardware architecture is further proposed to exploit the advantages of the optimized CNN model. We build a prototype system on field programmable gate array (FPGA) to validate and demonstrate the proposed design. We also map the proposed accelerator to 55nm CMOS technology. Experimental results show that the performance as well as the energy-resource efficiency of the proposed design are significantly higher than other existing vision-based HAR systems.
20 July 2023 in IEEE Transactions on Circuits and Systems I: Regular Papers
Neural Radiance Field (NeRF), a disruptive method for 3D representation and rendering, is extremely popular in the field of computer graphics and computer vision in the past three years. The most distinctive feature of NeRF models is their scene representation property, making it possible to quantize the models according to the complexity of the representing scenes. This paper proposes a novel approach to improve the efficiency of NeRF rendering by adopting precision-scalable computation. We first analyze and validate the idea of scene-dependent quantization for NeRF models. Based on that, we further propose look-up table (LUT) processing element (PE)-based precision-scalable computation unit designs. To evaluate the performance of different precision-scalable computing units, we implement these designs and compare the corresponding area, power, speed and energy efficiency. We also compare the proposed designs with existing approaches as well as the fixed precision approach for NeRF rendering tasks. The comparison results show that energy efficiency can be significantly improved by using precision-scalable computation for NeRF.
4 November 2025 in IEEE Transactions on Neural Networks and Learning Systems
Sparse-view computed tomography (SVCT) is an advancement in computed tomography (CT) technology that aims to reduce the radiation dose during imaging. Reconstructing high-quality images from sparse-view (SV) projections is an ill-posed inverse problem. Recently, implicit neural representations (INRs) as a self-supervised paradigm for solving underdetermined inverse problems have demonstrated excellent performance in SVCT reconstruction. However, since INR-based approaches rely on subject-specific training, they require a significant investment of time to optimize from scratch. Consequently, previous INR methods have not been able to meet the requisite timeliness of reconstruction. In our work, we propose RTSyner, an algorithm-hardware collaboration framework that facilitates the real-time efficiency of CT reconstruction. On the algorithmic side, we introduce an efficient coordinate-based feature module that exploits the local latent features as a positional external condition, leveraging the limited structural information of corrupted images derived from the sensory domain. By fusing latent features and coordinate information, the model learns a neural representation of the final tomographic image. On the hardware side, we design a dedicated hardware architecture with a customized algorithm flow to improve reconstruction speed and reduce power consumption. Furthermore, we improve the efficiency of model inference through model quantization, which also facilitates the subsequent deployment of hardware. Our extensive experimental results demonstrate that the RTSyner based on neural representation has achieved real-time SVCT reconstruction through the synergistic acceleration of the algorithm and hardware. We further explore its application potential via volume reconstructions under more complex acquisition geometries.
9 June 2026 in IEEE Transactions on Circuits and Systems I: Regular Papers
While high-resolution inputs improve recognition performance, they impose significant hardware overhead due to the computational intensity and memory bandwidth demands of conventional image signal processing (ISP). This inefficiency primarily stems from the structural mismatch between pixel-level ISP operations and the semantic information required by back-end neural networks. To address this, we propose an end-to-end RAW-based frequency-domain vision pipeline designed for efficient high-resolution detection. Under normal imaging conditions, the proposed pipeline bypasses the full human-perception-oriented ISP and directly processes RAW sensor data through lightweight channel mapping and a hardware-friendly approximate discrete cosine transform (DCT). To minimize front-end data movement, we reformulate the color-space transformation within the frequency domain and retain only a compact set of task-relevant low-frequency coefficients. Furthermore, we employ a multiplier-less, shift-and-add-based arithmetic architecture to simplify control logic and enhance energy efficiency. The pipeline is fully compatible with standard convolutional neural networks, requiring no architectural changes to the back-end detector. We validated the design using a Field-Programmable Gate Array (FPGA)-based prototype on the Xilinx ZCU102 platform integrated with a YOLOv3-Tiny detector. The system achieves real-time processing of $6144\times 4096$ RAW inputs at 33.18 frames per second (FPS). Experimental results demonstrate that the proposed frequency-domain approach maintains detection accuracy comparable to RGB-based pipelines while significantly reducing hardware resource utilization and power consumption.