论文部分内容阅读
提出了一种用于模式识别的新型多输入开关电流模糊处理器。该处理器是一个模数混合系统。其中，数字部分用于对输入码进行译码，并且产生系统所需的时钟和控制信号，模拟部分用于处理模糊信息并输出识别结果。该系统可以采用分时的方法接收未知模式的多个特征。模糊处理利用加权求和－求大操作，其中综合函数的权重可调，以提高处理器的自适应能力，并拓展其应用场合。侧重于集成电路的模拟程序ＰＳＰＩＣＥ模拟结果表明，该系统性能好、精度高。同时，电路在规模上易扩展。由于采用开关电流技术，该系统电路可以直接采用标准的数字互补型金属氧化物半导体（ＣＭＯＳ）工艺来实现，便于模、数混合集成，易于超大规模集成电路（ＶＬＳＩ）的制作。
A new multi-input switch current fuzzy processor for pattern recognition is proposed. The processor is an analog-to-digital hybrid system. The digital part is used to decode the input code and generate the clock and control signals required by the system. The analog part is used to process the fuzzy information and output the recognition result. The system can use time-sharing methods to receive multiple features of unknown patterns. Fuzzy processing uses weighted summing-seeking large operations, in which the weight of the integrated function is adjustable to improve the processor’s self-adaptability and to expand its application. The simulation of PSPIC, which focuses on the integrated circuit, shows that the system has good performance and high precision. At the same time, the circuit scales easily. Due to the switching current technique, the system circuit can be directly implemented by using a standard digital complementary metal-oxide-semiconductor (CMOS) process, which facilitates the integration of the analog and digital signals and facilitates the fabrication of very large scale integrated circuits (VLSI).