Yi-Chung Wu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Yi-Chung Wu has a total of 71 co-authors. Here are the top co-authors by number of shared publications:
- Chia-Hsiang Yang
National Taiwan University
8 shared publications - Jui-Hung Hung
Department of Computer Science, National Yang Ming Chiao Tung University, Hsinchu, Taiwan
5 shared publications Chia-Hsiang Yang
National Taiwan University, Graduate Institute of Electronics Engineering, Taipei, Taiwan
3 shared publicationsChao-Hsi Lee
GeneASIC Technologies Corporation, Hsinchu, Taiwan
2 shared publicationsChi-Shi Chen
Taiwan Semiconductor Research Institute, Hsinchu, Taiwan
2 shared publicationsChun-Pin Lin
Taiwan Semiconductor Research Institute, Hsinchu, Taiwan
2 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 73  (11) , 7691-7702
A High-Throughput FPGA Accelerator for Real-Time Sort-Free 3-D Gaussian Splatting
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 73  (11) , 7691-7702
A High-Throughput FPGA Accelerator for Real-Time Sort-Free 3-D Gaussian Splatting
- 27 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 19  (6) , 1105-1119
A 28nm Fully Integrated End-to-End Genome Analysis Accelerator for Next-Generation Sequencing
This paper presents the first end-to-end next-generation sequencing (NGS) data analysis accelerator for short-read mapping, haplotype calling, variant calling, and genotyping. It supports both single-end and paired-end short-reads (or reads) and uses the FM-index, a compact index data structure, for exact-match in short-read mapping. For inexact match part of short-read mapping, a dynamic programming array is proposed to determine the mapping results. To reduce the workload of short-read mapping, a rapid similarity calculation is designed. A rescue technique is also adopted to increase the overall sensitivity. In haplotype calling, a parallel k-mer processing engine can construct the de Bruijn graph and assemble the haplotypes. The variant calling step determines variants between a subject and a reference genome sequence with a variant discovery engine. Lastly, genotype likelihood is computed in parallel by a genotype likelihood computing engine, which outputs genotypes of all discovered variants and corresponding Phred-scaled likelihood (PL) values. This work completes end-to-end data analysis for the 50× PrecisionFDA dataset in an average of 28.2 minutes. It achieves a 3-to-59× higher throughput than the existing solutions with higher precision (99.79%) and sensitivity (99.03%). The chip also achieves a 935× higher energy efficiency than the Illumina DRAGEN FPGA acceleration system.
- 27 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 19  (6) , 1105-1119
A 28nm Fully Integrated End-to-End Genome Analysis Accelerator for Next-Generation Sequencing
This paper presents the first end-to-end next-generation sequencing (NGS) data analysis accelerator for short-read mapping, haplotype calling, variant calling, and genotyping. It supports both single-end and paired-end short-reads (or reads) and uses the FM-index, a compact index data structure, for exact-match in short-read mapping. For inexact match part of short-read mapping, a dynamic programming array is proposed to determine the mapping results. To reduce the workload of short-read mapping, a rapid similarity calculation is designed. A rescue technique is also adopted to increase the overall sensitivity. In haplotype calling, a parallel k-mer processing engine can construct the de Bruijn graph and assemble the haplotypes. The variant calling step determines variants between a subject and a reference genome sequence with a variant discovery engine. Lastly, genotype likelihood is computed in parallel by a genotype likelihood computing engine, which outputs genotypes of all discovered variants and corresponding Phred-scaled likelihood (PL) values. This work completes end-to-end data analysis for the 50× PrecisionFDA dataset in an average of 28.2 minutes. It achieves a 3-to-59× higher throughput than the existing solutions with higher precision (99.79%) and sensitivity (99.03%). The chip also achieves a 935× higher energy efficiency than the Illumina DRAGEN FPGA acceleration system.
- 18 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 60  (5) , 1818-1829
A 44.3-mW 62.4-fps Hyperspectral Image Processor for Spectral Unmixing in MAV Remote Sensing
This article presents the first dedicated processor designed to support the complete spectral unmixing workflow for hyperspectral image (HSI) processing, including rank reduction, endmember extraction, and abundance estimation. The design employs architecture explorations, including folding and data interleaving, to reduce hardware complexity. To enhance the throughput, the processor incorporates deeply pipelined reconfigurable processing elements (PEs) for compute-intensive tasks involved in spectral unmixing. The proposed sparsity-adaptive clocking technique leverages data sparsity and minimizes dynamic power consumption. Fabricated in a 40-nm CMOS technology, the proposed processor occupies a core area of 2.56 mm2. The chip consumes 44.3 mW of power at a clock frequency of 175 MHz from a 0.68-V supply. The processor can concurrently generate eight endmembers and their associated abundances for a ${256}{\times }{ 256}{\times }64$ HSI, resulting in a throughput of 62.4 fps. Comparative analysis with a high-end CPU demonstrates a significant processing speed improvement of $544{\times }$ , accompanied by energy efficiency that is 1735 $537{\times }$ higher and area efficiency that is 31 $647{\times }$ higher. The proposed processor is $17.5{\times }$ faster, with 236 $735{\times }$ higher energy efficiency and $4158{\times }$ higher area efficiency in comparison to a high-end graphics processing unit (GPU). The proposed processor provides a promising solution to support real-time hyperspectral remote sensing, particularly for battery-powered micro air vehicles (MAVs).
- 18 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 60  (5) , 1818-1829
A 44.3-mW 62.4-fps Hyperspectral Image Processor for Spectral Unmixing in MAV Remote Sensing
This article presents the first dedicated processor designed to support the complete spectral unmixing workflow for hyperspectral image (HSI) processing, including rank reduction, endmember extraction, and abundance estimation. The design employs architecture explorations, including folding and data interleaving, to reduce hardware complexity. To enhance the throughput, the processor incorporates deeply pipelined reconfigurable processing elements (PEs) for compute-intensive tasks involved in spectral unmixing. The proposed sparsity-adaptive clocking technique leverages data sparsity and minimizes dynamic power consumption. Fabricated in a 40-nm CMOS technology, the proposed processor occupies a core area of 2.56 mm2. The chip consumes 44.3 mW of power at a clock frequency of 175 MHz from a 0.68-V supply. The processor can concurrently generate eight endmembers and their associated abundances for a ${256}{\times }{ 256}{\times }64$ HSI, resulting in a throughput of 62.4 fps. Comparative analysis with a high-end CPU demonstrates a significant processing speed improvement of $544{\times }$ , accompanied by energy efficiency that is 1735 $537{\times }$ higher and area efficiency that is 31 $647{\times }$ higher. The proposed processor is $17.5{\times }$ faster, with 236 $735{\times }$ higher energy efficiency and $4158{\times }$ higher area efficiency in comparison to a high-end graphics processing unit (GPU). The proposed processor provides a promising solution to support real-time hyperspectral remote sensing, particularly for battery-powered micro air vehicles (MAVs).
- 26 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 56  (1) , 123-135
A 975-mW Fully Integrated Genetic Variant Discovery System-on-Chip in 28 nm for Next-Generation Sequencing
This article presents the first dedicated system-on-chip (SoC) that supports full data analysis workflow for genetic variant discovery for next-generation sequencing (NGS). The SoC implements four major steps: preprocessing, short-read mapping, haplotype calling, and variant calling. By adopting the sBWT-based short read mapping and Genome Analysis ToolKit haplotype caller-based variant calling algorithms, this work achieves a comparable precision as the software solutions. The dedicated hardware architecture achieves high parallelism with low hardware complexity. In this SoC, the multitask sorting engine and the dynamic programming processing engine are proposed for essential computing tasks for NGS data analysis. An ARC processor is integrated to facilitate IO interfaces, including a DDR3 PHY. Fabricated in a 28-nm technology, the chip area is 12 mm 2 . It dissipates 975 mW at a clock frequency of 400 MHz from a 0.9-V supply. The SoC is able to complete variant discovery for the whole human genome within 37 min. Compared with an optimized graphic processing unit solution, this work is 66$\times $ faster and also achieves an$1.59\times 10^{4}$ times higher energy efficiency and 3086$\times $ higher area efficiency. Verified with the FDA standardized benchmark, the SoC achieves a precision of 99.6%. The prototype system demonstrates the analysis procedure in real time for on-site DNA sequencing.- 26 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 56  (1) , 123-135
A 975-mW Fully Integrated Genetic Variant Discovery System-on-Chip in 28 nm for Next-Generation Sequencing
This article presents the first dedicated system-on-chip (SoC) that supports full data analysis workflow for genetic variant discovery for next-generation sequencing (NGS). The SoC implements four major steps: preprocessing, short-read mapping, haplotype calling, and variant calling. By adopting the sBWT-based short read mapping and Genome Analysis ToolKit haplotype caller-based variant calling algorithms, this work achieves a comparable precision as the software solutions. The dedicated hardware architecture achieves high parallelism with low hardware complexity. In this SoC, the multitask sorting engine and the dynamic programming processing engine are proposed for essential computing tasks for NGS data analysis. An ARC processor is integrated to facilitate IO interfaces, including a DDR3 PHY. Fabricated in a 28-nm technology, the chip area is 12 mm 2 . It dissipates 975 mW at a clock frequency of 400 MHz from a 0.9-V supply. The SoC is able to complete variant discovery for the whole human genome within 37 min. Compared with an optimized graphic processing unit solution, this work is 66$\times $ faster and also achieves an$1.59\times 10^{4}$ times higher energy efficiency and 3086$\times $ higher area efficiency. Verified with the FDA standardized benchmark, the SoC achieves a precision of 99.6%. The prototype system demonstrates the analysis procedure in real time for on-site DNA sequencing.- 23 August 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 54  (10) , 2907-2916
A 12.6 mW, 573–2901 kS/s Reconfigurable Processor for Reconstruction of Compressively Sensed Physiological Signals
This article presents a reconfigurable processor based on the alternating direction method of multipliers (ADMM) algorithm for reconstructing compressively sensed physiological signals. The architecture is flexible to support physiological ExG [electrocardiography (ECG), electromyography (EMG), and electroencephalography (EEG)] signal with various signal dimensions (128, 256, 384, and 512). Data characteristics are utilized to substantially reduce the overall hardware complexity by up to 99%. A 16x folded architecture achieves a 64% area-power product reduction compared with the unfolded one. A customized buffer is used for multi-word access, which reduces data latency by four times. It dissipates 75% less power with only 25% area when compared with the realization with conventional flip-flops. As a proof of concept, a reconfigurable processor for reconstructing ExG signals is presented. Fabricated in a 40-nm CMOS technology, the processor integrates 3.69-M gates in 3.23 mm². The chip delivers a throughput of 573-2901 kSamples/s (kS/s) for ExG signals and dissipates less than 12.6 mW at 87 MHz from a 0.60-V supply. Compared with state-of-the-art designs, the chip achieves a 1.5-to-14x higher throughput with 3.2-to-11x less energy, given the performance specification [reconstruction signal-to-noise ratio (RSNR) ≥ 15 dB].
- 23 August 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 54  (10) , 2907-2916
A 12.6 mW, 573–2901 kS/s Reconfigurable Processor for Reconstruction of Compressively Sensed Physiological Signals
This article presents a reconfigurable processor based on the alternating direction method of multipliers (ADMM) algorithm for reconstructing compressively sensed physiological signals. The architecture is flexible to support physiological ExG [electrocardiography (ECG), electromyography (EMG), and electroencephalography (EEG)] signal with various signal dimensions (128, 256, 384, and 512). Data characteristics are utilized to substantially reduce the overall hardware complexity by up to 99%. A 16x folded architecture achieves a 64% area-power product reduction compared with the unfolded one. A customized buffer is used for multi-word access, which reduces data latency by four times. It dissipates 75% less power with only 25% area when compared with the realization with conventional flip-flops. As a proof of concept, a reconfigurable processor for reconstructing ExG signals is presented. Fabricated in a 40-nm CMOS technology, the processor integrates 3.69-M gates in 3.23 mm². The chip delivers a throughput of 573-2901 kSamples/s (kS/s) for ExG signals and dissipates less than 12.6 mW at 87 MHz from a 0.60-V supply. Compared with state-of-the-art designs, the chip achieves a 1.5-to-14x higher throughput with 3.2-to-11x less energy, given the performance specification [reconstruction signal-to-noise ratio (RSNR) ≥ 15 dB].
- 25 February 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 66  (11) , 3156-3167
Design and In Vivo Verification of a CMOS Bone-Guided Cochlear Implant Microsystem
Objective: To develop and verify a CMOS bone-guided cochlear implant (BGCI) microsystem with electrodes placed on the bone surface of the cochlea and the outside of round window for treating high-frequency hearing loss. Methods: The BGCI microsystem consists of an external unit and an implanted unit. The external system-on-chip (SOC) is designed to process acoustic signals through an acquisition circuit and an acoustic DSP processor to generate stimulation patterns and commands that are transmitted to the implanted unit through a 13.56 MHz wireless power and bidirectional data telemetry. In the wireless power te-lemetry, a voltage doubler/tripler (2X/3X) active rectifier is used to enhance the power conversion efficiency and generate 2 V and 3 V output voltages. In the wireless data telemetry, phase-locked loop (PLL) based binary phase-shift keying (BPSK) and load-shift keying (LSK) modulators/demodulators are adopted for the down-link and uplink data through high-Q coils, respectively. The im-planted chip with 4-channel high-voltage-tolerant stimulator gen-erates biphasic stimulation currents up to 800 μA. Results: Electri-cal tests on the fabricated BGCI microsystem has been performed to verify the chip functions. The in-vivo animal tests in guinea pigs have shown the evoked third wave (Wave eIII) of electrically evoked auditory brainstem response (EABR) waveforms. It is ver-ified that auditory nerves can be successfully stimulated and acoustic hearing can be partially preserved. Conclusion and Significance: Different from traditional cochlear implants, the pro-posed BGCI microsystem is less invasive, preserves partially acoustic hearing, and provides an effective alternative to treating high-frequency hearing loss.
- 25 February 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 66  (11) , 3156-3167
Design and In Vivo Verification of a CMOS Bone-Guided Cochlear Implant Microsystem
Objective: To develop and verify a CMOS bone-guided cochlear implant (BGCI) microsystem with electrodes placed on the bone surface of the cochlea and the outside of round window for treating high-frequency hearing loss. Methods: The BGCI microsystem consists of an external unit and an implanted unit. The external system-on-chip (SOC) is designed to process acoustic signals through an acquisition circuit and an acoustic DSP processor to generate stimulation patterns and commands that are transmitted to the implanted unit through a 13.56 MHz wireless power and bidirectional data telemetry. In the wireless power te-lemetry, a voltage doubler/tripler (2X/3X) active rectifier is used to enhance the power conversion efficiency and generate 2 V and 3 V output voltages. In the wireless data telemetry, phase-locked loop (PLL) based binary phase-shift keying (BPSK) and load-shift keying (LSK) modulators/demodulators are adopted for the down-link and uplink data through high-Q coils, respectively. The im-planted chip with 4-channel high-voltage-tolerant stimulator gen-erates biphasic stimulation currents up to 800 μA. Results: Electri-cal tests on the fabricated BGCI microsystem has been performed to verify the chip functions. The in-vivo animal tests in guinea pigs have shown the evoked third wave (Wave eIII) of electrically evoked auditory brainstem response (EABR) waveforms. It is ver-ified that auditory nerves can be successfully stimulated and acoustic hearing can be partially preserved. Conclusion and Significance: Different from traditional cochlear implants, the pro-posed BGCI microsystem is less invasive, preserves partially acoustic hearing, and provides an effective alternative to treating high-frequency hearing loss.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 261-262
A 12.6MW 573-2,901KS/S Reconfigurable Processor for Reconstruction of Compressively-Sensed Phvsiological Signals
This work presents a reconfigurable processor based on the alternating direction method of multipliers (ADMM) algorithm for reconstructing compressively-sensed signals. The chip delivers a throughput of 573-to-2,901KS/s for reconstructing physiological signals. It dissipates 12.6mW at 87 MHz at 0.6V. Compared to the state-of-the-art designs, the chip achieves a 5.7-to-14x higher throughput with 5-to-11x lower energy for the target reconstruction SNR (RSNR) ≥ 15dB.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 261-262
A 12.6MW 573-2,901KS/S Reconfigurable Processor for Reconstruction of Compressively-Sensed Phvsiological Signals
This work presents a reconfigurable processor based on the alternating direction method of multipliers (ADMM) algorithm for reconstructing compressively-sensed signals. The chip delivers a throughput of 573-to-2,901KS/s for reconstructing physiological signals. It dissipates 12.6mW at 87 MHz at 0.6V. Compared to the state-of-the-art designs, the chip achieves a 5.7-to-14x higher throughput with 5-to-11x lower energy for the target reconstruction SNR (RSNR) ≥ 15dB.
- 3 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 11  (6) , 1216-1225
A 135-mW Fully Integrated Data Processor for Next-Generation Sequencing
Next-generation sequencing (NGS) enables high-throughput sequencing, in which short DNA fragments can be sequenced in a massively parallel fashion. However, the essential algorithm behind the succeeding NGS data analysis, DNA mapping, is still excessively time consuming. DNA mapping can be partitioned into two parts: suffix array (SA) sorting and backward searching. Dedicated hardware designs for the less-complex backward searching have been proposed, but feasible hardware for the most complicated part, SA sorting, has never been explored. Based on the memory-efficient sBWT algorithm, this work is the first integrated NGS data processor for the entire DNA mapping. The $k$-ordered Ferragina and Manzini index used in the sBWT algorithm is leveraged to improve storage capacity and reduce hardware complexity. The proposed NGS data processor realizes the sBWT algorithm through bucket sorting, suffix grouping, and suffix sorting circuits. Key design parameters are analyzed to achieve the optimal performance with respect to hardware cost and execution time. Fabricated in 40-nm CMOS, the NGS data processor dissipates 135 mW at 200 MHz from a 0.9-V supply. With 1-GB external memory, the chip can analyze human DNA within 10 min. This work achieves 43 065$\times$ and 8 971$\times$ [3208$\times$ and 402$\times$] higher energy efficiency (throughput-to-area ratio) than the high-end CPU and GPU solutions, respectively.- 3 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Circuits and Systems
- Vol. 11  (6) , 1216-1225
A 135-mW Fully Integrated Data Processor for Next-Generation Sequencing
Next-generation sequencing (NGS) enables high-throughput sequencing, in which short DNA fragments can be sequenced in a massively parallel fashion. However, the essential algorithm behind the succeeding NGS data analysis, DNA mapping, is still excessively time consuming. DNA mapping can be partitioned into two parts: suffix array (SA) sorting and backward searching. Dedicated hardware designs for the less-complex backward searching have been proposed, but feasible hardware for the most complicated part, SA sorting, has never been explored. Based on the memory-efficient sBWT algorithm, this work is the first integrated NGS data processor for the entire DNA mapping. The $k$-ordered Ferragina and Manzini index used in the sBWT algorithm is leveraged to improve storage capacity and reduce hardware complexity. The proposed NGS data processor realizes the sBWT algorithm through bucket sorting, suffix grouping, and suffix sorting circuits. Key design parameters are analyzed to achieve the optimal performance with respect to hardware cost and execution time. Fabricated in 40-nm CMOS, the NGS data processor dissipates 135 mW at 200 MHz from a 0.9-V supply. With 1-GB external memory, the chip can analyze human DNA within 10 min. This work achieves 43 065$\times$ and 8 971$\times$ [3208$\times$ and 402$\times$] higher energy efficiency (throughput-to-area ratio) than the high-end CPU and GPU solutions, respectively.- 1 May 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Integration of energy-recycling logic and wireless power transfer for ultra-low-power implantables
This paper presents an integration of energy-recycling logic circuits with a wireless power transfer receiving module for ultra-low-power applications, such as transcutaneous biomedical implantables. In the prototype design, one inductive coil implanted inside the body receives wireless power and supplies the following electronics. While part of the loading is composed of conventional CMOS logics, the rest is implemented with energy-recycling logic circuits. Energy-recycling logic and the associated adiabatic operation achieve excellent energy efficiency by transferring and recycling energy between digital logic blocks along with the signal propagation. The required AC supplies further lead to a natural integration with wireless power transfer and therefore obviate the need for a rectifier that contributes to substantial power loss. As a proof of concept, a finite-impulse-response filter is designed in 90-nm CMOS process. Simulation results show a 59.3% power reduction as compared to static CMOS counterpart.
- 1 May 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Integration of energy-recycling logic and wireless power transfer for ultra-low-power implantables
This paper presents an integration of energy-recycling logic circuits with a wireless power transfer receiving module for ultra-low-power applications, such as transcutaneous biomedical implantables. In the prototype design, one inductive coil implanted inside the body receives wireless power and supplies the following electronics. While part of the loading is composed of conventional CMOS logics, the rest is implemented with energy-recycling logic circuits. Energy-recycling logic and the associated adiabatic operation achieve excellent energy efficiency by transferring and recycling energy between digital logic blocks along with the signal propagation. The required AC supplies further lead to a natural integration with wireless power transfer and therefore obviate the need for a rectifier that contributes to substantial power loss. As a proof of concept, a finite-impulse-response filter is designed in 90-nm CMOS process. Simulation results show a 59.3% power reduction as compared to static CMOS counterpart.