Frenger Paul
Working Hypothesis, Inc., P.O. Box 820506, Houston, TX 77282-0506, USA.
Biomed Sci Instrum. 2003;39:89-92.
This paper updates the author's 1999 RMBS presentation on digitally controlled linear artificial neuron design. Each neuron is based on a standard operational amplifier having excitatory and inhibitory inputs, variable gain, an amplified linear analog output and an adjustable threshold comparator for digital output. This design employs a 1-wire serial network of digitally controlled potentiometers and resistors whose resistance values are set and read back under microprocessor supervision. This system embodies several unique and useful features, including: enhanced neuronal stability, dynamic reconfigurability and network extensibility. This artificial neuronal is being employed for feature extraction and pattern recognition in an advanced robotic application.
本文更新了作者1999年关于数字控制线性人工神经元设计的RMBS报告。每个神经元基于一个标准运算放大器，具有兴奋性和抑制性输入、可变增益、放大的线性模拟输出以及用于数字输出的可调阈值比较器。该设计采用了一个由数字控制电位器和电阻组成的单线串行网络，其电阻值在微处理器的监控下进行设置和读取。该系统具有几个独特且有用的特性，包括：增强的神经元稳定性、动态可重构性和网络可扩展性。这种人工神经元正被用于先进机器人应用中的特征提取和模式识别。
