程泽军, 李彬鸿, 李博, 罗家俊, 韩郑生. 一种基于数字信号控制的CMOS神经元电路[J]. 微电子学与计算机, 2019, 36(1): 6-10, 15.
引用本文: 程泽军, 李彬鸿, 李博, 罗家俊, 韩郑生. 一种基于数字信号控制的CMOS神经元电路[J]. 微电子学与计算机, 2019, 36(1): 6-10, 15.
CHENG Ze-jun, LI Bin-hong, LI Bo, LUO Jia-jun, HAN Zheng-sheng. A Digital-Controlled Silicon Neuron Circuit[J]. Microelectronics & Computer, 2019, 36(1): 6-10, 15.
Citation: CHENG Ze-jun, LI Bin-hong, LI Bo, LUO Jia-jun, HAN Zheng-sheng. A Digital-Controlled Silicon Neuron Circuit[J]. Microelectronics & Computer, 2019, 36(1): 6-10, 15.

一种基于数字信号控制的CMOS神经元电路

A Digital-Controlled Silicon Neuron Circuit

  • 摘要: 采用0.18 μm CMOS工艺设计了一种由数字信号控制的神经元电路.相比于传统神经元电路, 本电路不需要偏置电压, 控制简单, 便于大规模集成.在数字信号的控制下, 本电路可以方便地实现Regular Spiking (RS), Fast Spiking (FS), Chattering (CH), Intrinsic Bursting (IB)四种神经元脉冲响应模式.此外本文介绍了一种激励型突触电路, 该电路的突触权值可以通过基于脉冲的Spike Driven Synaptic Plasticity(SDSP)学习机制进行调节.在此基础上本文利用3个神经元与2个突触实现了Pavlov实验, 证明了所设计的神经元电路支持SDSP学习规则, 可以用于构建神经形态硬件.

     

    Abstract: A digital controlled neuron circuit which mimics properties of biological neuron is presented using 0.18 μm CMOS technology. Compared with other neuron circuits, the proposed one does not require a bias voltage, which facilitates circuit control and is beneficial to large-scale integration. Under the control of digital signals, the proposed circuit can easily generate four types of firing patterns: Regular Spiking (RS), Fast Spiking (FS), Chattering (CH), and Intrinsic Bursting (IB). In addition, we design an excitatory synapse circuit whose synaptic weight can be adjusted according to Spike Driven Synaptic Plasticity (SDSP) learning algorithm. Furthermore, we implement Pavlov experiment with three neurons and two synapses. The experimental result demonstrates that this neuron circuit supports SDSP learning algorithm and can be used in neuromorphic hardware.

     

/

返回文章
返回