Department of Electrical and Computer Engineering, Sungkyunkwan University, Suwon 16416, Korea.
SKAIChips, Sungkyunkwan University, Suwon 16419, Korea.
Sensors (Basel). 2022 Mar 23;22(7):2459. doi: 10.3390/s22072459.
This paper presents a register-transistor level (RTL) based convolutional neural network (CNN) for biosensor applications. Biosensor-based diseases detection by DNA identification using biosensors is currently needed. We proposed a synthesizable RTL-based CNN architecture for this purpose. The adopted technique of parallel computation of multiplication and accumulation (MAC) approach optimizes the hardware overhead by significantly reducing the arithmetic calculation and achieves instant results. While multiplier bank sharing throughout the convolutional operation with fully connected operation significantly reduces the implementation area. The CNN model is trained in MATLAB on MNIST handwritten dataset. For validation, the image pixel array from MNIST handwritten dataset is applied on proposed RTL-based CNN architecture for biosensor applications in ModelSim. The consistency is checked with multiple test samples and 92% accuracy is achieved. The proposed idea is implemented in 28 nm CMOS technology. It occupies 9.986 mm of the total area. The power requirement is 2.93 W from 1.8 V supply. The total time taken is 8.6538 ms.
本文提出了一种基于寄存器-晶体管级 (RTL) 的卷积神经网络 (CNN)，用于生物传感器应用。目前需要使用生物传感器通过 DNA 识别来进行基于生物传感器的疾病检测。为此，我们提出了一种可综合的基于 RTL 的 CNN 架构。所采用的乘法和累加 (MAC) 并行计算技术通过显著减少算术计算来优化硬件开销，并实现即时结果。同时，通过在卷积运算中共享乘法器银行并结合全连接运算，显著减少了实现面积。该 CNN 模型在 MATLAB 上基于 MNIST 手写数据集进行训练。为了验证，将 MNIST 手写数据集的图像像素数组应用于基于 RTL 的生物传感器应用的 CNN 架构，在 ModelSim 中进行验证。通过多个测试样本进行一致性检查，实现了 92%的准确率。该想法在 28nm CMOS 技术中实现。它占据了总区域的 9.986 平方毫米。在 1.8V 电源下，功率需求为 2.93W。总时间为 8.6538ms。
