Dejian Li

Dejian Li

Auto-generated profile, based on publicly available information

Dejian Li

Beijing Smartchip Microelectronics Technology co., LTD., China

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 September 2023 in IEEE Embedded Systems Letters
Verified
Journal Article
Closed access
DLDejian Li
Dejian Li
XFXi Feng
Xi Feng
CSChongfei Shen
Chongfei Shen
QCQi Chen
Qi Chen
LYLixin Yang
Lixin Yang
MLMENG LIU
MENG LIU
This paper introduces a dedicated processor architecture, called MEGACORE, which leverages vector technology to enhance tracking performance in Visual Simultaneous Localization and Mapping (VSLAM) systems. By harnessing the inherent parallelism of vector processing and incorporating a floating point unit (FPU), MEGACORE achieves significant acceleration in the tracking task of VSLAM. Through careful optimizations, we achieved notable improvements compared to the baseline design. Our optimizations resulted in a 14.9% reduction in the area parameter and a 4.4% reduction in power consumption. Furthermore, by conducting application benchmarks, we determined that the average speedup ratio across all stages of the tracking process is 3.25. These findings highlight the effectiveness of MEGACORE in improving the efficiency and performance of VSLAM systems, making it a promising solution for real-world implementations in embedded systems.

28 May 2021
Verified
Conference Paper
Closed access
WHWang Hui
Wang Hui
Dongming WangDongming Wang
Dongming Wang
LDLi Dejian
Li Dejian
ZLZeng Lin
Zeng Lin
ZWZhe Wang
Zhe Wang
Anomaly detection is the primary method of detecting intrusion. Unsupervised models, such as auto-encoders network, auto-encoder, and GMM, are currently the most widely used anomaly detection techniques. In reality, the samples used to train the unsupervised model may not be pure enough and may include some abnormal samples. However, the classification effect is poor since these approaches do not completely understand the association between reconstruction errors, reconstruction characteristics, and irregular sample density distribution. This paper proposes a novel intrusion detection system architecture that includes data collection, processing, and feature extraction by integrating data reconstruction features, reconstruction errors, auto-encoder parameters, and GMM. Our system outperforms other unsupervised learning-based detection approaches in terms of accuracy, recall, F1-score, and other assessment metrics after training and testing on multiple intrusion detection data sets.

1 January 2026 in IEEE Access
Verified
Journal Article
Open access
LWLi Wang
Li Wang
KSKe Yang Sun
Ke Yang Sun
ZWZhe Wang
Zhe Wang
JGJie Gan
Jie Gan
YCYu Zhe Chen
Yu Zhe Chen
DLDe Jian Li
De Jian Li

20 July 2026 in ACM Transactions on Reconfigurable Technology and Systems
Verified
Journal Article
Open access
YSYuhao Sun
Yuhao Sun
DLDejian Li
Dejian Li
BWBin Wei
Bin Wei
BWBei Wang
Bei Wang
CWCHENGHUA WANG
CHENGHUA WANG
WLWeiqiang Liu
Weiqiang Liu
SPHINCS+ is the only hash-based post-quantum digital signature scheme standardized by NIST. However, its complex multi-level structure and hash-intensive computations pose significant challenges for high-performance and efficient hardware implementations. This paper presents a high-throughput and efficient hardware architecture for SPHINCS+, supporting three security levels. Exploiting the computational characteristics of SPHINCS+ hash functions, a pipelined hash engine and an iterative hash engine are co-designed to efficiently serve different hashing workloads. The architecture further incorporates bubble-free scheduling and parallel Winternitz One-Time Signature Plus (WOTS+) node computation, thereby maximizing the utilization of the pipelined hash engine and improving resource efficiency. In addition, a merged-computation scheme for Forest of Random Subsets (FORS) trees is proposed to reduce the hardware overhead of supporting multiple security levels. Compared with the state-of-the-art design on UltraScale+, the proposed architecture improves area-time product (ATP) by \(1.6\times/1.8\times\) for key generation and signing, and achieves \(3.8\times/4.2\times/1.9\times\) faster key generation, signing, and verification. Compared with the fastest existing implementation on Artix-7, it delivers \(1.2\times\) faster signing and verification, together with \(1.4\times\) better ATP.

10 August 2022
Verified
Conference Paper
Closed access
SHShunfeng Han
Shunfeng Han
DLDejian Li
Dejian Li
XFXi Feng
Xi Feng
GYGuan Yuan
Guan Yuan
BLBofu Li
Bofu Li
FLFengman Liu
Fengman Liu
In this paper, Chip-Package interaction(CPI) stress during chip attach reflow process were analyzed by FEM. Furthermore, there is a much distinct scales difference between 15mm×15mm FCCSP package and extreme low-k (ELK) chip, submodeling techniques had been used in this study to bridge the scale gap. Firstly, in order to get more accurate CPI stress in the submodel, it’s necessary to locate the weakest bump. A FEM thermal-mechanical stress analysis with global package model was conducted. In this global model, the PSV and PI layer weren’t established. The result showed the maximum stress of package appeared on the left lower side of the chip. Secondly, according to the simulation results of the global model, a 1/4 model of the package was established to reduce simulation time. Then, according to the 1/4 model simulation results, we establish submodel, which the PSV opening size is 40um, the PI opening size is 30um, and the bump size is 45um×70um, the chip thickness is 150um, and then the submodel simulation was carried out. So far, the detailed CPI stress was obtained. Finally, 4 structural parameters were chosen to conduct DOE analysis evaluating the effects on CPI stress, including of chip thickness, PSV opening size, PI opening size and bump size. The results showed that the stress of the ELK layer was directly concerned with the structural characters. By optimization on structure properties of the FCCSP package, the stress of the ELK layer can be reduced to the relatively safe level.

18 October 2024
Verified
Conference Paper
Open access
PGPo Gao
Po Gao
YZYanxin Zhang
Yanxin Zhang
LTLang Tan
Lang Tan
LYLixin Yang
Lixin Yang
BNBin Niu
Bin Niu
ZWZhihua Wang
Zhihua Wang
The RISC-V multi-core heterogeneous SOC has five harts that divided into three clusters, and integrates an eFPGA. Inter-cluster is asymmetric-multiprocessor and intra-cluster is symmetric-multiprocessor. The SoC not only has the powerful computing ability of general CPU, but also has the scalability and parallel processing ability of FPGA. But it also makes the process of loading and launching the firmware extremely complicated. According to the hardware architecture of SoC and possible use scenarios, four basic bare-metal and operating system combination boot schemes and three methods to load eFPGA executable files are designed. The actual test shows that the boot process can work as expected. These boot schemes cover almost all the usage scenario of SoC, expand the application range of SoC, and are more conducive to the popularization and use of SoC.