Düşük güçlü çok seviyeli CMOS sınıflandırıcı devresi
İnsanların günlük yaşamında belirli bir sesi, görüntüyü veya analog bir veriyi tanımak için kullandıkları kuralları tanımlamak oldukça karmaşık bir dizi işlem gerektirmektedir ve hatta bu kuralları tanımlamak bazen mümkün olamamaktadır. Oysa pratikte karşılaşılan örüntü tanıma olaylarını, yazılım ve donanım tabanlı tanıma uygulamalarında belirli kriterlere oturtmak mümkündür. Sınıflandırma yöntemleri ilk olarak örüntü sınıflandırma adı altında görülmeye başlanmış ve ilk algoritmalarda basit yapılar ele alınmıştır; ilk gerçeklenen yapıda en yakın komşu yakınsaması kullanılmıştır. Sınıflandırma işlemi, benzer özellik taşıyan objelerin başka farklı özellikte olanlardan ayırt edilmesi şeklinde tanımlanabilir ve otomatik hedef belirleme, yapay zekâ, yapay sinir ağları, analog-sayısal dönüştürücüler, kuantalama, tıbbi tanı, istatistik gibi çeşitli alanlarda kullanılır. Dolaysıyla da, günümüzde, gerek gerçek dünyada gerekse sayısal dünyada verilerin sınıflandırılması büyük önem taşımaktadır. Bugüne kadar sınıflandırma işlemi genellikle çeşitli algoritmalar yardımıyla yazılımsal olarak yapılmaktaydı, oysa birçok uygulamada, sınıflandırma işlemini daha hızlı ve gerçek zamanda yapmak gerektiğinden bu algoritmaların donanımsal olarak gerçeklenmeleri çok daha yararlı olmaktadır. Ayrıca günümüzde portatif cihazların da artmasından dolayı donanımsal olarak gerçeklenecek cihazlarda da güç tüketimi büyük önem kazanmıştır. Dolayısıyla sınıflandırıcı devrelerin de bu ihtiyaçları karşılayacak şekilde tasarlanması gerekmektedir. Bu makalede akım-modlu düşük güçte çalışan bir sınıflandırıcı devresi sunulmaktadır. Önerilen sınıflandırıcı devresi, temel bir bloktan yararlanmaktadır; bu temel bloklar kullanılarak daha gelişmiş sınıflandırıcı yapılarının gerçekleştirilebileceği gösterilmiştir. Önerilen devrenin benzetimleri için 0.35 μm AMS CMOS teknoloji parametreleri kullanılmıştır. Ayrıca çekirdek devre adı verilen temel bloğun, tek boyutlu ve iki boyutlu sınıflandırıcı yapılarının benzetim sonuçları verilmiştir.
A low-power multilevel CMOS classifier circuit
In the everyday life of humans, to define the rules used to recognize a certain sound, image or an analog data necessitates a sequence of complex processes which sometimes becomes impossible to accomplish. However, to develop well defined software and hardware based criteria in the application of pattern recognition problems, is possible. The aim of classification can be defined as to assign an unknown object to a class containing similar objects (or to distinguish objects having the same properties from those not possessing). Classification is especially important in the real world applications or in the digital world. Basic classification methods using nearest neighbourhood algorithm have first been seen in early sixties under the subject tile” pattern recognition.” Classification is used in a huge variety of applications such as automatic target identification, artificial neural networks, artificial intelligence, template matching, pattern recognition, analog to digital converters, quantization, medical diagnosis, statistics etc. Therefore nowadays, be it in the real or digital world, data classification is becoming increasingly important. But until recently, major work on classification was on developing algorithms used in software packages whereas, in many applications it is becoming more and more important to classify data much faster and in real time, entailing the need for hardware realization of these algorithms. Software approaches are not practical for real time applications, the processing is computationally very expensive, consuming a lot of Central Processing Unit (CPU) time when implemented as software running on general purpose computers. So in literature hardware implementation of classifier topologies become necessary. Also in literature hardware realized classifiers which are designed to work in low power operation; moreover some of these hardware classifiers do not have custom tunability. So they can only be used for a specific application. The recent developments in electronics technology has created a perfect medium for the hardware realization of classifier structures which, in turn, will render many classifier application prospects feasible in real time. This paper targets the design and application to real world problems of tunable, low power new classifier circuits using CMOS technology. So, a low-power CMOS implementation of a multi-input data classifier with several output levels is presented. The proposed circuit operates in current-mode and can classify several types of analog vector data. An architecture is developed comprising a threshold circuit based on CMOS transistors operating in subthreshold region. To this purpose a one dimensional classifier, called core circuit is proposed. The core circuit also works as a one-dimensional classifier. As this circuit is designed to operate in currentmode the input and the output data is provided to the core circuit with currents. So by interconnecting several core circuits and adding the output currents a multi output classifier can be obtained. Also, combining several core circuits in groups in such a way that each group has identical input current (different from the others), a multi-dimensional, multi-level output classifier can be obtained. Also, numerous efforts in balancing the trade off between power consumption, area and speed have resulted in an acceptable performance. On the other hand, the rapid increasing use of battery operated portable equipment in application areas such as telecommunications and medical electronics increases the importance of low-power and small sized VLSI circuits’ technologies. One solution to achieve lowpower and acceptable performance is to operate the transistors in the subthreshold region. The CMOS transistors working in subthreshold region are suitable only for specific applications which need, not very high performance, but low power consumption. The primary aim of this paper is to develop a low power classifier circuit with n inputs and externally tunable decision regions with different output amplitude for each region. Due to the subthreshold operation of the transistors in the proposed core circuit, very low power consumption becomes possible. The proposed core circuit is constructed with two threshold and a subtractor circuit. The SPICE simulation of the threshold circuit, core circuit, one dimensional and two dimensional classifier circuits are given. Using 0.35 μm parameters of AMS CMOS technology, SPICE simulations are performed and a low-power, custom tunable classifier circuit is realized. Because of the parallel processing characteristic of the circuit, it is well suited for real-world applications.
___
- Abdel-Aty-Zohdy H.S. ve Al-Nsour M., (1999). Reinforcement learning neural network circuits for electronic nose, IEEE International Symposium on Circuits and Systems, 5, 30 May-2 June, 379 – 382, Orlando, FL.
- Geske, G., Stupmann, F. ve Wego, A., (2003). High Speed Color Recognition With an Analog Neural Network Chip, IEEE International Conference on Industrial Technology, 1, Dec 10-12, 104-107, Maribor, Slovenia.
- Hodges D., Jackson H. ve Saleh R., (2004). Analysis and Design of Digital Integrated Circuits, Third edition, Mc Graw Hill, 7-64, USA.
- Hunt E., (1975). Artificial Intelligence. New York Academic Press, 4-46, USA.
- Lin G. ve Shi B., (1999a). A current-mode sorting circuit for pattern recognition, Intelligent Processing and Manufacturing of Materials, July 10-15, 1003 – 1007, Hawaii.
- Lin G. ve Shi B., (1999b). A multi-input currentmode fuzzy integrated circuit for pattern Recognition, Second International Conference on IntelligentProcessing and Manufacturing Materials, July 10-15, 687-693, Hawaii.
- Liu B., Chen C. ve Tsao J.,(2000). A Modular Current-Mode Classifier Circuit for Template Matching Application, IEEE Trans. on Circuit and Systems-II, Analog and Digital Signal Processing, 47, 2, 145-151.
- Morgül A. ve Temel T., (2005). Curent-mode level restoration circuit for multi-valued logic, Electronics Letters, 41, 5, 230-231.
- Yıldız M., Minaei S. ve Göknar C., (2006a). CMOS Realization of a Quantized-Output Classifier Circuit, 13th IEEE International Conference on Electronics, Circuits and Systems, Dec 10 - 13, 292-295, France.
- Yıldız M., Minaei S. ve Göknar C., (2006b). Current Mode Double Threshold Neuron Activation Function, Complex Computing-Networks: Brainlike and Wave-oriented Electrodynamic Algorithms. Springer Proceedings in Physics, 104, 267-274.
- Yıldız M., Minaei S. ve Göknar C., (2007a). A Low- Power Multilevel-Output Classifier Circuit, European Conference on Circuit Theory and Design, August 26-30, 747-750, Spain.
- Yıldız M., Minaei S. ve Göknar C., (2007b). A CMOS Classifier Circuit using Neural Networks with Novel Architecture, IEEE Transaction on Neural Networks, 18, 6, 1845-1849.