Data processing: artificial intelligence – Fuzzy logic hardware – Having function generator
Reexamination Certificate
1998-08-21
2001-03-20
Powell, Mark (Department: 2122)
Data processing: artificial intelligence
Fuzzy logic hardware
Having function generator
C706S001000, C706S004000
Reexamination Certificate
active
06205438
ABSTRACT:
BACKGROUND OF THE INVENTTON
1. Field of the Invention
The present invention relates to a fuzzy processor, more particularly, to a hybrid current-type fuzzy processor which is of high precision, suitable in many applications, and manufactured with low cost.
2. Description of Related Art
The idea of fuzzy logic was introduced by L. A. Zadeh in 1965, which plays an important role in the field of computer science and has been successfully applied in many applications. Fuzzy logic is commonly implemented in computer software. However, because software fuzzy logic can not satisfy many applications requiring real-time processing, a design of hardware fuzzy logic has become an important research approach.
Currently, two types of hardware fuzzy logic are provided, digital hardware fuzzy logic and analog hardware fuzzy logic. The digital hardware fuzzy logic is supported by lots of well-known digital VLSI technologies but it suffers a disadvantage in having complicated circuitry. In addition, because the inputs to fuzzy logic are usually analog signals, A/D and D/A converters are required for the digital hardware fuzzy logic, which makes the structure of the fuzzy logic more complicated. On the other hand, the analog hardware fuzzy logic is easy to implement due to the non-linear characteristic of the analog circuitry. Moreover, no A/D or D/A converter is required and thus the structure of the fuzzy logic is simple.
Generally, analog fuzzy logic is constructed by multi-value logic circuit units, which may be of a voltage type or a current type. For conventional voltage type circuits, operational amplifiers are required for summation or subtraction operations to voltages, which makes the circuit complicated. However, the current type circuit is capable of proceeding summation and subtraction operations to currents and thus simplifies the circuit. In addition, the operating speed of a current type circuit is generally higher than that of the voltage type circuit because the gain bandwidth of the operational amplifier restricts the operating speed of the voltage type circuit. Moreover, in a voltage type fuzzy logic circuit, switch capacitors are usually required, which increases the size of a chip for the circuit because a large chip area is required to fabricate a capacitor. The use of switch capacitors also increases the complexity of manufacturing a chip for the circuit as two polysilicon layers are required for fabricating a switch capacitor. However, the fabrication of a current switch for the current type fuzzy logic can be done by standard digital CMOS technology and thus reduce the complexity of manufacturing a chip for the circuit. Accordingly, the present invention provides a switch current type fuzzy processor for pattern recognition.
Because of the imprecise, vague and incomplete nature of available information about an object, people usually have to proceed in fuzzy inference and adjudgement, which forms the physical basis for the fuzzy mathematics applied in pattern identification. The principle rule of fuzzy pattern identification is the maximum membership degree rule. In the actual application, a standard pattern usually has a plurality of fuzzy features. If there are n standard patterns each having m fuzzy features and the jth fuzzy characteristic of the ith pattern is Aij, where i=1,2, . . . , n; j=1,2, . . . , m, then each standard pattern Ai is a fuzzy vector Ai=<Ai1, Ai2, . . . , Aim>, 1 ≦i≦n. Assuming that u=(u
1
, u
2
, . . . , u
m
) is a pattern to be identified, each member of u, that is each u
j
, corresponds to a fuzzy characteristic. If there exists an i &egr; {1,2, . . . , n} such that &mgr;
Ai
(u)=max{&mgr;
A1
(u), &mgr;
A2
(u), . . . , &mgr;
An
(u)} then u relatively belongs to Ai, wherein it is assumed that &mgr;
Ai
(u)=M
m
(&mgr;
Ai1(u
1
), &mgr;
Ai2(u
2
), . . . , &mgr;
Aim
(u
m
)) and M
m
() is a synthesis function.
There are many choices for the synthesis functions. The most frequently used are the minimum-finding function, &Lgr;
j
(X)=&Lgr;
j
x
j
for j=1 to m, and the weighting summation function, &Sgr;X=&Sgr;&agr;
j
x
j
for j=1 to m. The minimum-finding finding function is not suitable for pattern identification since it only emphasizes a local characteristic and neglects the other features. The weighting summation function is able to emphasize a local characteristic by adjusting the weighting factor &agr;
j
while not neglecting the other features whereby it is suitable for pattern identification. Therefore, the weighting summation function is adopted by the fuzzy processor in accordance with the present invention. Meanwhile, to enhance the ability of self-adjustment thereby increasing the application fields, the weighting factors of the synthesis function is designed to be adjustable.
SUMMARY OF THE INVENTON
The general objective of the present invention is to provide a current-type fuzzy processor for pattern identification, which is of high precision, suitable in many applications, and manufactured with low cost. To attain this, the fuzzy processor comprises a characteristic decoder for decoding a to-be-identified pattern, a plurality of membership function generators for generating a plurality of membership degrees for the to-be-identified pattern, a weighting decoder for generating a weighting value, a plurality of weighting adjustment circuits for performing weighting operations on the plurality of membership degrees according to the weighting value to generate a plurality of weighted membership degrees, a plurality of accumulators for performing accumulation operations on the plurality of weighted membership degrees to generate a plurality of synthesis membership degrees, a winner-take-all circuit for outputting the plurality of synthesis membership degrees in an order of magnitude; and a timing and control logic for generating timing and control signals for the fuzzy processor.
REFERENCES:
patent: 4694418 (1987-09-01), Ueno et al.
patent: 5295226 (1994-03-01), Yamakawa
patent: 5396579 (1995-03-01), Koji et al.
patent: 5706497 (1998-01-01), Takahashi et al.
patent: 5719999 (1998-02-01), Nishidai et al.
patent: 5799291 (1998-08-01), Eichfeld et al.
patent: 5802204 (1998-09-01), Basehore
patent: 5802506 (1998-10-01), Hutchison
patent: 5918221 (1999-06-01), Manaresi et al.
patent: 6038556 (2000-03-01), Hutchison
Ramierz-Angulo, “Current-Mode and Voltage-Mode VLSI Fuzzy Processor Architecture,” May 1995 IEEE International Symposium on Circuits and Systems, vol. 2, pp. 1156-1159.*
Chao-Lieh, “Fuzzy Hardware Synthesis with Generic LR Fuzzy Cells,” Proceedings of the Second New Zealand International Two-Stream Conference on Artifical Neural Networks and Expert Systems, 1995, pp. 128-131, Nov. 1995.
Lin Gu
Shi Bingxue
Kolsich Hartwell Dickinson McCormack & Heuser
Ponder Michael
Powell Mark
Winbond Electronics Corp.
LandOfFree
Current-type fuzzy processor does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Current-type fuzzy processor, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Current-type fuzzy processor will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2549506