Recognition unit and recognition apparatus

• US 5,542,005 A
• Filed: 03/13/1995
• Issued: 07/30/1996
• Est. Priority Date: 09/17/1991
• Status: Expired due to Fees

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Sakaue, Shigeo, Yamamoto, Hiroshi, Kojima, Yoshihiro, Maruno, Susumu, Shimeki, Yasuharu, Kohda, Toshiyuki
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): KELLEY, CHRISTOPHER S
• Application Number:
  US08/403,126
• Time in Patent Office: 505 Days
• Field of Search: 382/155, 382/158, 395/21, 395/23
• US Class Current: 382/155
• CPC Class Codes: G06F 18/24317 Piecewise classification, i.e. whereby each classification requires several discriminant rules
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology

Abstract

A recognition apparatus is provided with a plurality of recognition units organized in a multilayered hierarchical structure. Each of the recognition units includes a signal input section, a quantizer for performing a quantization according to a signal inputted from the signal input section, and a path selecting section for performing a selection of paths according to an output from the quantizer. The path selecting section includes a path input section having at least one path input terminal, a path output section having at least one path output terminal, a load distribution selecting section for selecting a load distribution, and a load setting section for changing the strength of connection between the path input terminal and the path output terminal according to the output of the quantizer by the use of the load distribution selected by the load distribution selecting section.

Claims

1. A recognition apparatus having a network organized in a multilayered hierarchical structure comprising:

a plurality of first recognition units each comprising;

a first signal input section;

a first quantizer for performing a quantization according to a signal inputted from said first signal input section; and

a first path selecting section for performing a selection of paths according to an output from said first quantizer, said first path selecting section comprising;

a first path input section having only one path input terminal;

a first path output section having a plurality of first path output terminals for outputting respective output signals;

a load distribution selecting section for selecting a load distribution;

a load setting section for changing a strength of connection between said first path input terminal and said plurality of first path output terminals according to the output of said first quantizer by use of the load distribution selected by said load distribution selecting section; and

wherein said load distribution selected by said load distribution selecting section is centered at a level position determined by the quantization and has a maximum value other than "0" at the level position determined by the quantization, and has another value less than or equal to said maximum value but greater than "0" at a position close to the level position determined by the quantization, and has a value of "0" at other positions;

an operation mode judging section for judging as to whether a learning mode or a recognition mode is selected, wherein said load distribution selecting section is controlled by said operation mode judging section; and

a plurality of second recognition units, each of which is connected with at least one of said first recognition units and comprises;

a second path input section having a plurality of second path input terminals for receiving said output signals from said first path output section of said at least one of said first recognition units and an adder for adding signals inputted thereto via said path input terminals; and

a path output section having only one path output terminal and a threshold processor for performing a threshold processing with respect to an output signal from said adder;

wherein unless a desired one of said second recognition units provides an output greater than any other of said second recognition units in the learning mode, said operation mode judging section controls said load distribution selecting section to alter the load distribution by adding a first predetermined value to each of current load values.

2. The apparatus according to claim 1 further comprising an operation mode judging section for judging as to whether an initial learning mode or an additional learning mode is selected, wherein said load distribution selecting section is controlled by said operation mode judging section.

3. The apparatus according to claim 2, wherein the load distribution selected in the additional learning mode has a spread narrower than that of the load distribution selected in the initial learning mode.

4. The apparatus according to claim 1 further comprising a learning progress judgment section for judging a learning progress, wherein said load distribution selecting section is controlled by said learning progress judgment section.

5. The apparatus according to claim 4, wherein as the learning advances, the spread of the load distribution becomes narrow.

6. The apparatus according to claim 1, wherein the load distribution selected in the recognition mode has a spread narrower than that of the load distribution selected in the learning mode.

14. The apparatus according to claim 1, wherein said first predetermined value is obtained by multiplying each of the current load values by a second predetermined value.

7. A recognition apparatus having a network organized in a multilayered hierarchical structure comprising:

a plurality of first recognition units each comprising;

a first signal input section;

a first quantizer for performing a quantization according to a signal inputted from said first signal input section; and

a first path selecting section for performing a selection of paths according to an output from said first quantizer, said first path selecting section comprising;

a first path input section having only one first path input terminal;

a first path output section having a plurality of first path output terminals for outputting respective output signals;

a load distribution selecting section for selecting a load distribution;

a first load setting section for changing a strength of connection between said first path input terminal and said plurality of first path output terminals according to the output of said first quantizer by use of the load distribution selected by said load distribution selecting section; and

wherein said load distribution selected by said load distribution selecting section is centered at a level position determined by the quantization and has a maximum value other than "0" at the level position determined by the quantization, and has another value less than or equal to said maximum value but greater than "0" at a position close to the level position determined by the quantization, and has a value of "0" at other positions;

a plurality of second recognition units, each of which is connected with at least one of said first recognition units and comprises;

a second signal input section;

a second quantizer for performing a quantization according to a signal inputted from said second signal input section; and

a second path selecting section for performing a selection of paths according to an output from said second quantizer, said second path selecting section comprising;

a second path input section having only one second path input terminal for receiving a respective output signal from a first path output terminal of said first path output section of said at least one of said first recognition units;

a second path output section having a plurality of second path output terminals for outputting respective output signals;

a load calculating section adapted for calculating a load distribution;

a second load setting section for changing a strength of connection between said second path input terminal and said second path output terminal according to the output of said second quantizer by use of an output from said load calculating section; and

wherein said load distribution calculated by said load calculating section is centered at a level position determined by the quantization and has a maximum value other than "0" at the level position determined by the quantization, and has another value less than or equal to said maximum value but greater than "0" at a position close to the level position determined by the quantization, and has a value of "0" at other positions; and

a plurality of third recognition units, each of which is connected with at least one of said second recognition units and comprises;

a third path input section having a plurality of third path input terminals for receiving the output signals from said second path output terminals of said second path output section of said at least one of said second recognition units and an adder for adding signals inputted thereto via said third path input terminals; and

a third path output section having only one third path output terminal and a threshold processor for performing a threshold processing with respect to an output signal from said adder.

8. The apparatus according to claim 7 further comprising an operation mode judging section for judging as to whether an initial learning mode or an additional learning mode is selected, wherein said load distribution selecting section is controlled by said operation mode judging section.

9. The apparatus according to claim 8, wherein the load distribution selected in the additional learning mode has a spread narrower than that of the load distribution selected in the initial learning mode.

10. The apparatus according to claim 7 further comprising a learning progress judgment section for judging a learning progress, wherein said load distribution selecting section is controlled by said learning progress judgment section.

11. The apparatus according to claim 10, wherein as the learning advances, the spread of the load distribution becomes narrow.

12. The apparatus according to claim 7 further comprising an operation mode judging section for judging as to whether a learning mode or a recognition mode is selected, wherein said load distribution selecting section is controlled by said operation mode judging section.

13. The apparatus according to claim 12, wherein the load distribution selected in the recognition mode has a spread narrower than that of the load distribution selected in the learning mode.

15. The apparatus according to claim 12, wherein said load calculating section calculates a new load distribution when a desired one of said third recognition units does not provide an output greater than any other of said third recognition units in the learning mode.

16. The apparatus according to claim 12 further comprising an error judging section for calculating an amount of error between a teacher signal and an output from a network having a plurality of recognition units organized in a multilayered hierarchical structure wherein the load distribution is calculated by said load calculating section according to the amount of error calculated by said error judging section.