神经系统中随机共振现象研究

【作者】 韩晓鹏；

【导师】 楼正国； 钱鸣奇；

【作者基本信息】 浙江大学 ， 生物医学工程， 2005， 硕士

【摘要】 长期来人们对于噪声总以为是有害的,可是作用于非线性系统的噪声在一定条件下却是有益的,这在神经系统中表现尤为突出。本文就是讨论神经系统中噪声的作用。更确切地说是讨论神经系统中的随机共振现象。 随机共振是一种噪声增强系统响应的现象,当弱信号输入非线性系统时,噪声的引入可以增强系统对输入信号的响应。随机共振可以作为神经系统在噪声环境中检测微弱信号的一种机制,本文以Hodgkin-Huxley神经元模型为基础,在神经元的输入中引入噪声,通过仿真计算的方法研究了单个神经元和双层神经元网络中的随机共振现象。 低于神经元阈值的信号输入单个神经元时,神经元的输出信噪比以及输出与输入之间的互信息率在一个非零噪声强度上达到最大值,这表明了随机共振现象的存在。进一步研究发现,单个神经元对阈下输入信号的检测性能随噪声强度或阈下信号幅度的改变而变化,因而它不具备在噪声强度不稳的环境中检测幅度多变的输入信号的能力。接着,模拟生物神经系统中神经元之间会聚的连接方式,构建了一个双层神经元网络,研究结果表明该网络可以发生随机共振现象,而且当噪声强度或者阈下信号幅度在一定范围内变化时,该网络都可以对输入信号进行很好地检测,这说明双层神经元网络的信号检测性能要优于单今神经元,另一方面也说明了神经元之间会聚的连接方式在神经系统信号检測方面具有重要意义。最后研究了阈上信号作用下神经元响应的情况,当输入信号幅度高于神经元的阈值时,神经元的响应在一些时候并不能很好地反映输入信号的变化,研究结果表明,加入一定强度的噪声,可以改善神经元对输入信号的响应,神经元的输出信噪比可以在一个非零噪声强度上达到最大值,这个结果将发生随机共振的信号幅度范围从阈值以下扩展到了阈值以上。更多还原

【Abstract】 Noise is usually considered a nuisance. But in certain nonlinear systems, the presence of noise can actually enhance the response of a system to weak input signals This phenomenon, called stochastic resonance , was frequently reported to be discovered in the nervous system. Here we performed a simulation study of stochastic resonance in neuronal systems.Firstly, stochastic resonance in a single Hodgkin-Huxley neuron was characterized with signal-to-noise ratio and mutual information rate for showing the positive role of noise in subthreshold signal detection. But the ability of a single neuron in delecting subthreshold signal changes as the noise intensity or the amplitude of the subthreshold signal varies. This imposes a limitation on a single neuron in detecting signals with varying amplitude when the background noise intensity is unstable. After studying stochastic resonance in a single neuron, a double layer Hodgkin-Huxley neuronal network was constructed for studying the ability of the network in subthreshold signal detection. The network was investigated and the characteristic of stochastic resonance was observed. Results indicate that the ability of the network in detecting the subthreshold signal changes little when the noise intensity or the amplitude of the subthreshold signal varies within a certain range. Thus, the network does not suffer from, to a certain extent, the limitation in subthreshold signal detection for a single neuron. Intuitively, a neuron will response well to suprathreshold input and adding noise will degrade the neuron’s response. In fact, it is not always the situation. Our results show that noise can enhance the response of a neuron to some suprathreshold signals. And stochastic resonance for the suprathreshold signal was observed.更多还原