ZYZenghui Yu

Zenghui Yu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Zenghui Yu has a total of 18 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    SSKD: Stepwise Self-Knowledge Distillation for Binary Neural Networks in Keyword Spotting

    by Hailong Zou, Jionghao Zhang, Jun Li, Hang Ran, Wulve Yang, Rui Zhou, Zenghui Yu, Yi Zhan, Shushan Qiao

    The hardware power-aware keyword spotting (KWS) implementation requires small memory footprint, low-complex computation, and high accuracy performances. Binary neural networks (BNNs) naturally satisfy these constraints. They quantize both weights and activations to 1-bit. This reduces storage and replaces most multiply–accumulate operations with bitwise operations. However, such extreme quantization incurs substantial information loss and leaves a noticeable accuracy gap relative to full-precision models. Optimization is also more difficult because the sign function is non-differentiable, and surrogate-gradient updates introduce gradient mismatch. To preserve the hardware benefits of BNNs while alleviating the accuracy degeneration induced by 1-bit quantization, this article addresses the problem from two complementary aspects: Firstly, a Stepwise Self-Knowledge Distillation (SSKD) training approach is proposed to effectively improve the student BNN’s accuracy performance. The SSKD training framework achieves effective supervision for student BNNs. A Stepwise Training Strategy is proposed to optimize the training stability and accuracy. Weight Scaling Factor improves the student’s representational capability. Secondly, an extremely lightweight Binary Temporal Convolutional ResNet (BTC-ResNet) is also proposed. The parameters and calculations inside the network are greatly reduced for the inference. Experiments on the GSCD v1 and GSCD v2 benchmarks demonstrate the effectiveness of our methods for low-power keyword spotting. For the 12-class task, BTC-ResNet14 achieves 97.23% accuracy on GSCD v1 and 97.31% on GSCD v2 with 0.75 Mb parameters and 1.35 M FLOPs. For the 35-class task on GSCD v2, it reaches 95.56% accuracy with 0.76 Mb parameters and 1.35 M FLOPs. These results indicate that our method achieves a competitive accuracy–efficiency balance relative to recent distillation-based BNN KWS baselines reported in the comparative experiments. All these studies are helpful and promising for future KWS deployment on low-power hardware devices.

  • Open Access

    Hardware Implementation of Block Floating-Point FFT Based on Approximate Computation and Conflict-Free Access

    by Xiaohui Zhang, Zenghui Yu, Shushan Qiao, Yi Zhan

    This paper proposes a novel Fast Fourier Transform hardware architecture for natural-order multi-input multi-output processing, which is based on approximate computing and the block floating-point data format. The architecture innovatively combines the low-power characteristics of approximate multipliers with the high-precision advantages of block floating-point data formats, enabling efficient processing of 2048-point and 4096-point Fast Fourier Transform data. By employing conflict-free memory access techniques, the design optimizes data scheduling, reduces idle cycles during processing, and significantly enhances energy efficiency while maintaining computational accuracy. Experimental validation demonstrates that the processor, fabricated using a 40nm process, achieves a compact chip area of 0.42 mm². Operating at a frequency of 671.14 MHz, it delivers a throughput of 1073.82 MS/s for 4096-point FFT computations, with a power consumption of only 110.85 mW and a signal-quantization-to-noise ratio of 58.51 dB. Compared to state-of-the-art designs, this architecture achieves breakthrough improvements in throughput performance while maintaining competitive energy efficiency.


  • Open Access

    An Efficient Codebook Search Algorithm for Line Spectrum Frequency (LSF) Vector Quantization in Speech Codec

    by Yuqun Xue, Yongsen Wang, Jianhua Jiang, Zenghui Yu, Yi Zhan, Xiaohua Fan, Shushan Qiao

    A high-performance vector quantization (VQ) codebook search algorithm is proposed in this paper. VQ is an important data compression technique that has been widely applied to speech, image, and video compression. However, the process of the codebook search demands a high computational load. To solve this issue, a novel algorithm that consists of training and encoding procedures is proposed. In the training procedure, a training speech dataset was used to build the squared-error distortion look-up table for each subspace. In the encoding procedure, firstly, an input vector was quickly assigned to a search subspace. Secondly, the candidate code word group was obtained by employing the triangular inequality elimination (TIE) equation. Finally, a partial distortion elimination technique was employed to reduce the number of multiplications. The proposed method reduced the number of searches and computation load significantly, especially when the input vectors were uncorrelated. The experimental results show that the proposed algorithm provides a computational saving (CS) of up to 85% in the full search algorithm, up to 76% in the TIE algorithm, and up to 63% in the iterative TIE algorithm. Further, the proposed method provides CS and load reduction of up to 29–33% and 67–69%, respectively, over the BSS-ITIE algorithm.


  • Open Access

    A Low-Power High-Speed Sense-Amplifier-Based Flip-Flop in 55 nm MTCMOS

    by Heng You, Jia Yuan, Weidi Tang, Zenghui Yu, Shushan Qiao

    In this paper, a sense-amplifier-based flip-flop (SAFF) suitable for low-power high-speed operation is proposed. With the employment of a new sense-amplifier stage as well as a new single-ended latch stage, the power and delay of the flip-flop is greatly reduced. A conditional cut-off strategy is applied to the latch to achieve glitch-free and contention-free operation. Furthermore, the proposed SAFF can provide low voltage operation by adopting MTCMOS optimization. Post-layout simulation results based on a SMIC 55 nm MTCMOS show that the proposed SAFF achieves a 41.3% reduction in the CK-to-Q delay and a 36.99% reduction in power (25% input data toggle rate) compared with the conventional SAFF. Additionally, the delay and the power are smaller than those of the master-slave flip-flop (MSFF). The power-delay-product of the proposed SAFF shows 2.7× and 3.55× improvements compared with the conventional SAFF and MSFF, respectively. The area of the proposed flip-flop is 8.12 μm2 (5.8 μm × 1.4 μm), similar to that of the conventional SAFF. With the employment of MTCMOS optimization, the proposed SAFF could provide robust operation even at supply voltages as low as 0.4 V.


  • Open Access

    Fast Computation of LSP Frequencies Using the Bairstow Method

    by Yuqun Xue, Zhijiu Zhu, Jianhua Jiang, Yi Zhan, Zenghui Yu, Xiaohua Fan, Shushan Qiao

    Linear prediction is the kernel technology in speech processing. It has been widely applied in speech recognition, synthesis, and coding, and can efficiently and correctly represent the speech frequency spectrum with only a few parameters. Line Spectrum Pairs (LSPs) frequencies, as an alternative representation of Linear Predictive Coding (LPC), have the advantages of good quantization accuracy and low spectral sensitivity. However, computing the LSPs frequencies takes a long time. To address this issue, a fast computation algorithm, based on the Bairstow method for computing LSPs frequencies from linear prediction coefficients, is proposed in this paper. The algorithm process first transforms the symmetric and antisymmetric polynomial to general polynomial, then extracts the polynomial roots. Associated with the short-term stationary property of speech signal, an adaptive initial method is applied to reduce the average iteration numbers by 26%, as compared to the statics in the initial method, with a Perceptual Evaluation of Speech Quality (PESQ) score reaching 3.46. Experimental results show that the proposed method can extract the polynomial roots efficiently and accurately with significantly reduced computation complexity. Compared to previous works, the proposed method is 17 times faster than Tschirnhus Transform, and has a 22% PESQ improvement on the Birge-Vieta method with an almost comparable computation time.