Data communication method and apparatus using neural-networks

• US 4,972,473 A
• Filed: 07/20/1989
• Issued: 11/20/1990
• Est. Priority Date: 07/22/1988
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Inouchi, Hidenori, Iwamura, Kazuaki, Ejiri, Masakazu, Kakumoto, Shigeru
• Primary Examiner(s): Buczinski, Stephen C.
• Assistant Examiner(s): Gregory, Bernarr Earl
• Application Number:
  US07/382,198
• Time in Patent Office: 488 Days
• Field of Search: 380/3-7, 380/9, 380/10, 380/20, 380/49, 380/50, 364/200, 364/900, 364/274.9, 364/276.6, 364/917.91, 364/917.95, 364/513, 364/602, 364/807
• US Class Current: 380/243
• CPC Class Codes: G06N 3/02 Neural networks
• CPC Class Codes: H04N 1/448 Rendering the image unintelligible, e.g. scrambling

Abstract

A data communication apparatus comprises: means for dividing data to be transmitted into a plurality of blocks and extracting the data from each block; a first multi-layered neural network of three or more layers which has weighting coefficients to output the same data as the input data for the data extracted from each block and which can output data from an intermediate layer; the transmission data extracted from each block being inputted to the first neural network and outputted from the intermediate layer; means for encoding the transmission data which is outputted from the intermediate layer of the first neural network and, thereafter, transmitting; means for receiving and decoding the transmitted data; a second multi-layered neural network of three or more layers which has the same weight coefficients as those of the first neural network and can input data from an intermediate layer; the decoded data of each block being inputted to the second neural network and outputted from an output layer; and means for restoring the data on the basis of the output data from the output layer of the second neural network.

Claims

1. A data communication apparatus using a neural network, comprising:

image memory means for storing image data to be transmitted;

image block forming means for dividing said image data into a plurality of blocks, each block having a predetermined size, and for extracting the divided image data from each block;

image converting means for executing a predetermined image converting process on the extracted image data of each block to produce converted image data;

a first multi-layered neural network for encrypting data and having at least an input layer, an intermediate layer and an output layer, and which can output encrypted data from said intermediate layer in response to the application of data to said input layer, said first multi-layered neural network having predetermined weighting coefficients such that the same image data as applied to the input layer is outputted from the output layer;

means for inputting the converted image data which was converted in each block by said image converting means to the input layer of said first multi-layered neural network and for outputting encrypted image data from the intermediate layer of said first multi-layered neural network;

coding means for encoding encrypted image data outputted from the intermediate layer of the first multi-layered neural network and for encoding converting information regarding the converting process which is executed by said image converting means, and for transmitting the encoded encrypted image data and the converting information;

decoding means for receiving and decoding the encoded encrypted image data and converting information transmitted from said coding means;

a second multi-layered neural network for decrypting data and having at least an input layer, an intermediate layer and an output layer, and which can output decrypted data from said output layer in response to the application of encrypted data to said intermediate layer, said second multi-layered neural network having predetermined weighting coefficients such that the same image data as applied to the input layer is outputted from the output layer;

means for inputting the encrypted image data of each block which is provided by said decoding means to the intermediate layer of said second multi-layered neural network and for outputting image data from the output layer of said second multi-layered neural network; and

image converting means for converting the image data of each block which is outputted from the output layer of said second multi-layered neural network in accordance with said received converting information.

2. An apparatus according to claim 1, wherein said predetermined image converting process of said image converting means is a process to rearrange an arrangement of pixels forming each block into a predetermined sequence.

3. An apparatus according to claim 1, wherein said predetermined image converting process of said image converting means is a process to invert information of pixels forming each block.

4. An apparatus according to claim 1, wherein said predetermined image converting process of said image converting means is a process to add a predetermined value to information of pixels forming each block.

5. An apparatus according to claim 1, wherein in said first neural network, coefficients to couple special neuron elements forming the intermediate layer of the first neural network and special neuron elements constructing forming the output layer coupled with said neuron elements of the intermediate layer are forcedly set to a predetermined value.

6. An apparatus according to claim 5, wherein said coefficients are set to 0.

7. An apparatus according to claim 1, wherein the weighting coefficients of said second multi-layered neural network are the same as the weighting coefficients of said first muli-layered neural network.

8. A data communication method using a neural-network, comprising the steps of:

(1) dividing data to be transmitted into a plurality of blocks, each block having a predetermined size, and extracting said data from each block;

(2) inputting the extracted data to be transmitted to a first multi-layered neural network for encrypting data and having at least an input layer, an intermediate layer and an output layer, and which can output encrypted data from said intermediate layer in response to the application of data to said input layer, said first multi-layered neural network having predetermined weighting coefficients such that the same data as applied to the input layer is outputted from the output layer and outputting encrypted data from the intermediate layer;

(3) encoding the encrypted data which is outputted from the intermediate layer of said first multi-layered neural network and, thereafter, transmitting said encoded encrypted data;

(4) receiving and decoding the transmitted encoded encrypted data to produce encrypted data; and

(5) inputting the encrypted data of each block to an intermediate layer of a second multi-layered neural network for decrypting data and which has the same construction and the same weighting coefficients as those of the first multi-layered neural network and in which encrypted data is inputted to the intermediate layer and outputted from the output layer.

9. A method according to claim 8, further having the steps of:

(6) executing a predetermined data converting process on the image data of each block which was extracted in said first step; and

(7) converting and restoring the data of each block which was outputted from the output layer of said second multi-layered neural network in said fifth step in accordance with information regarding the predetermined data converting process executed in said sixth step.

10. A method according to claim 9, wherein the information regarding the predetermined data converting process executed in said sixth step is transmitted prior to transmission of the encrypted data in said third step.

11. A method according to claim 8, wherein the weighting coefficients of said second neural network which are used in said fifth step are transmitted prior to transmission of the encrypted data in said third step.

12. A method according to claim 8, wherein weighting coefficients to couple specific neuron elements forming the intermediate layer of said first neural network which are used in said second step and neuron elements forming the output layer coupled with said neuron elements of the intermediate layer of said first multi-layered neural network are forcedly set to a predetermined value.

13. A method according to claim 12, wherein said coefficients are set to 0 in both neural networks.

14. A method according to claim 8, wherein the data to be transmitted is image data.

15. A method according to claim 8, wherein the data to be transmitted is a voice signal.

16. A data transmission apparatus using a neural network, comprising:

data block forming means for dividing data to be transmitted into a plurality of blocks, each block having a predetermined size, and for extracting the data from each block;

a multi-layered neural network for encrypting data and having at least an input layer, an intermediate layer and an output layer, and which can output encrypted data from said intermediate layer in response to the application of data to said input layer, said first multi-layered neural network having predetermined weighting coefficients such that the same image data as applied to the input layer is outputted from the output layer;

means for inputting data of each block extracted by said data block forming means to the input layer of said multi-layered neural network and for outputting encrypted data from said intermediate layer of said multi-layered neural network; and

coding means for encoding the encrypted data outputted from the intermediate layer of said multi-layered neural network and for transmitting the encoded encrypted data.

17. An apparatus according to claim 16, further having data converting means for executing a predetermined data converting process on the data of each block extracted by said data block forming means;

and wherein the data of each block converted by said data converting means is input to the input layer of said multi-layered neural network.

18. An apparatus according to claim 16, wherein coefficients coupling specific neuron elements forming the intermediate layer of said neural network and special neuron elements forming the output layer coupled with said specific neuron elements of the intermediate layer are set to a predetermined value.

19. A system for communicating data, such as an image, comprising:

a transmitting unit including means for dividing data to be transmitted into a plurality of blocks, a first neural network for encrypting data and which comprises at least an input layer, an intermediate layer and an output layer and which can output data from said intermediate layer which has predetermined weighting coefficients such that the same data as inputted to said input layer of said first neural network from each of said block is outputted said output layer thereof, and means for outputting data from said intermediate layer in an unit of said block; and

a receiving unit including a second neural network for decrypting data and having at least an input layer, an intermediate layer and an output layer, means for directly inputting said data transmitted to said intermediate layer of said second neural network which has the same weighting coefficients as those of said first neural network, means for decoding said data transmitted in said second neural network to provide reproduced data and means for storing the reproduced data in a unit of said block.

20. A system according to claim 19, wherein said transmitting unit further includes a converter which receives a data block from said dividing means and changes an original arrangement of the data in a block in accordance with predetermined sequence numbers, the output of said converter being inputted to the first neural network;

and said receiving unit further includes a converter to recover the original arrangement of the data in the block obtained as an output of the second neural network in accordance with said predetermined sequence numbers.

21. A system according to claim 19, wherein said transmitting unit further includes control means for allowing the blocks to be extracted and transmitted in accordance with predetermined sequence numbers, and said receiving unit further includes control means for rearranging the reproduced data of each block in accordance with said predetermined sequence numbers.

22. A data communication system using neural networks, comprising:

a transmitting apparatus having;

(1) means for storing data;

(2) a first neural network for encrypting data and comprising at least an input layer, an intermediate layer and an output layer, and having weighting coefficients predetermined by learning so that data output from said output layer is the same data as data received by said input layer, wherein said input layer is connected to said storing means, and first data applied to said input layer is encrypted and outputted from said intermediate layer as second data; and

a receiving apparatus for receiving data transmitted from said transmitting apparatus having;

(3) a second neural network for decrypting data comprising at least an input layer, an intermediate layer and an output layer, having weighting coefficients determined on the basis of said weighting coefficients of said first neural network, wherein third data accepted by said intermediate layer of said second neural network is decrypted and outputted from said output layer of said second neural network as fourth data.

23. An apparatus according to claim 22, wherein said weighting coefficients of said first neural network are transmitted as control data from said transmitting apparatus.

24. An apparatus according to claim 22, wherein said weighting coefficients of said first neural network are transmitted together with said second data as control data from said transmitting apparatus.

25. An apparatus according to claim 22, wherein said transmitting apparatus further comprises control means for dividing said data stored in said storing means into a plurality of blocks and for reading out said data from each of said blocks.

26. An apparatus according to claim 25, wherein said transmitting apparatus further comprises converting means connected between said storing means and said input layer of said first neural network for performing a conversion predetermined upon data in blocks which are read out from said storing means and for providing said converted data to said input layer of said first neural network as said first data.

27. An apparatus according to claim 26, wherein said converting means comprises means for inverting an order of data in said blocks read out from said storing means.

28. An apparatus according to claim 26, wherein said converting means comprises means for adding a certain value to said data in said blocks read out from said storing means.

29. An apparatus according to claim 22, wherein said weighting coefficients predetermined by learning are coefficients which connect neuron elements constituting said intermediate layer with neuron elements constituting said output layer in said first neural network.

30. An apparatus according to claim 29, wherein at least one of said weighting coefficients is set to a predetermined value in independence on said learning.

31. An apparatus according to claim 30, wherein said predetermined value is zero.

32. An apparatus according to claim 22, wherein said transmitting apparatus comprises means for encoding said second data and for transmitting encoded second data.

33. An apparatus according to claim 32, wherein said receiving apparatus comprises means for decoding signals received from said transmitting apparatus and for providing decoded signals to said intermediate layer of said second neural network.

34. An apparatus according to claim 22, wherein said data stored in said storing means is image data.

35. An apparatus according to claim 22, wherein said data stored in said storing means is audio signal data.

36. A data communication apparatus comprising:

means for storing data;

a first neural network for encrypting data and comprising at least an input layer, an intermediate layer and an output layer, and having weighting coefficients predetermined so that data output from said output layer is the same as data received by said input layer, wherein first data is inputted to said input layer and second data is outputted from said intermediate layer in the form of encrypted data;

means for encoding said second data and for transmitting encoded second data as third data;

means for receiving said third data and for decoding encoded third data to produce fourth data; and

a second neural network for decrypting data and comprising at least an input layer, an intermediate layer and an output layer, having weighting coefficients determined on the basis of said weighting coefficients of said first neural network, wherein said fourth data is inputted to said intermediate layer of said second neural network and fifth data in the form of decrypted data is outputted from said output layer of said second neural network.

37. An apparatus according to claim 36, wherein said second data is encoded and transmitted with information concerning said weighted coefficients of said first neural network.

38. An apparatus according to claim 36, wherein said data stored in said storing means is divided into a plurality of blocks and read out from each of said blocks to said first neural network.

39. An apparatus according to claim 36, further comprising:

converting means for changing an order of data which is read out from said storing means and for outputting said converted data as said first data.

40. An apparatus according to claim 36, further comprising:

converting means for inverting an order of data read out from said storing means to form converted data and for outputting said inverted data as said first data.

41. An apparatus according to claim 36, further comprising:

converting means for adding a predetermined value to data read out from said storing means to form converted data and for outputting said inverted data as said first data.

42. An apparatus according to claim 36, wherein said weighting coefficients are coefficients which connect neuron elements constituting said intermediate layer with neuron elements constituting said output layer in said first neural network.

43. An apparatus according to claim 42, wherein at least one of said weighting coefficients is set to a predetermined value.

44. An apparatus according to claim 43, wherein said predetermined value is zero.

45. An apparatus according to claim 36, wherein said data stored in said storing means is image data.

46. An apparatus according to claim 36, wherein said data stored in said storing means is audio signal data.

47. A data communication method, comprising:

(A) in a transmitting process performed by a transmitting apparatus comprising a first neural network for encrypting data and which is provided with an input layer, an intermediate layer and an output layer and which has weighting coefficients predetermined so that data output from said output layer is the same as data received by said input layer, the steps of;

(a1) applying first data to be transmitted to said input layer of said first neural network;

(a2) outputting encrypted first data as second data from said intermediate layer; and

(a3) transmitting said second data;

(B) in a receiving process performed by a receiving apparatus comprising a second neural network for decrypting data and which is provided with an input layer, an intermediate layer and an output layer and which has weighting coefficients determined on the basis of said weighting coefficients of said first neural network, the steps of;

(b1) applying second data received from said transmitting process to said intermediate layer of said second neural network; and

(b2) outputting decoded data as third data from said output layer of said second neural network.

48. A method according to claim 47, comprising in said transmitting process the steps of:

dividing data stored in a storing means into a plurality of blocks;

reading out said stored data from each of said blocks; and

performing a predetermined conversion upon said read out data.

49. A method according to claim 47, wherein said first data is image data.

50. A method according to claim 47, wherein said first data in audio signal data.