摘要
从生物神经元的电化学特性出发,基于积分发放(I&F)电路理论模型,提出了一种新型的结构紧凑的脉冲编码CMOS神经元电路,模仿神经元细胞体输出连续脉冲串。该模型的优点在于大大简化了模型结构,其运行结果很好地拟合了神经元的生理特性,且在工艺参数不可调节的情况下,可通过输入信号灵活控制电路结构,改变输入耦合权重,从而实现对输入信号的脉冲编码。HSPICE仿真结果表明,该电路可以通过输出脉冲串频率实现对多端输入的二进制方波信号的权重识别,在自适应耦合调整的信息传递,图像识别神经网络构建和信号调制方面具有很大的应用前景。
According to the electrochemical characteristics of biological neurons,this paper presents a new compact pulse coded CMOS neuron circuit based on IF model to imitate a continuous pulse steam generated by neuron cell.We show that the model has been simplified greatly,it characterizes many aspects of real neurons.In the case that the fabrication parameters of transistors cannot be adjusted,the circuit construction and coupled weights can be controlled flexibly by the input signals,then the circuit realizes the pulse code modulation of the input signals.The results of HSPICE simulation show that this circuit can realize the weight identification of the binary square-wave input signal according to the frequency of output pulse stream.This new pulse coded CMOS neuron circuit would have great application prospects in the information transmission,the construction of image recognition neural network and signal modulation.
出处
《电子器件》
CAS
2011年第3期286-291,共6页
Chinese Journal of Electron Devices
基金
国家863计划项目(2007AA03Z303)
国家973计划课题(2010CB934104)
