Method and apparatus for automatic segregation and routing of signals of different origins by using prototypes

• US 6,246,985 B1
• Filed: 08/20/1998
• Issued: 06/12/2001
• Est. Priority Date: 08/20/1998
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Kanevsky, Dimitri, Maes, Stephane H., Zlatsin, Alexander, Zadrozny, Wlodek Wlodzimierz
• Primary Examiner(s): {haeck over (S)}mits, Ta̅livaldis I.
• Application Number:
  US09/137,966
• Time in Patent Office: 1,027 Days
• Field of Search: 704/256, 704/270, 379/93.11, 379/211, 455/31.2, 370/351
• US Class Current: 704/270
• CPC Class Codes: H04L 45/00 Routing or path finding of packets in data switching networks routing or path finding in wireless networks H04W40/00
• CPC Class Codes: H04M 2203/2066 Call type detection of indication, e.g. voice or fax, mobile of fixed, PSTN or IP
• CPC Class Codes: H04M 3/42 Systems providing special services or facilities to subscribers specially adapted for wireless communication networks H04W4/00

Abstract

A method and apparatus is disclosed for automatic segregation of signals of different origin, using models that statistically characterize a wave signal, more particularly including feature vectors consisting of a plurality of parameters extracted from a data stream of a known type for use in identifying data types by comparison, which can be Hidden Markov Model based methods, thereby enabling automatic data type identification and routing of received data streams to the appropriate destination device, thereby further enabling a user to transmit different data types over the same communication channel without changing communication settings.

Claims

1. An apparatus for segregation of signals of different origin, comprising:

one or more computer systems at a receiving end, each having a central processing unit (CPU) and a memory, and each being connected to a network;

one or more communication devices that send or receive a plurality of data via said network, said data being of a plurality of origins and combined into a common data stream at said receiving end;

means, executed by said CPU at a receiving end, for processing allowable data types residing on said one or more computer systems, and for creating one or more data models;

means, executed by said CPU, using said data models for parsing said data stream into its component segments, each said segment having one of said plurality of origins, and for determining a destination device for each said segment, and for sending each said segment to its corresponding destination device.

2. The apparatus of claim 1, wherein said destination devices can be any one or more of the following:

telephone, fax machine, printer, card reader, biometrics reader, pager, PC, laptop, calculator, speaker, mike, TV, video camera, radio, audio player, different devices that process audio, different devices that process texts, or different devices that process graphics.

3. The apparatus of claim 2, wherein said receiving end is located in a car and wherein said destination devices can additionally include one or more of the following:

map reader, audio web player, or car radar system.

4. The apparatus of claim 1, wherein said plurality of data origins can be any one or more of the following:

voice data, non-voice data, speech, music, textual data, or graphics.

5. The apparatus of claim 1, wherein said means for parsing and determining further comprise:

means for identifying one or more of said segments by comparing said data stream with a set of known protocols stored at said receiving end;

if a segment is identified with a known protocol, means for further identifying said segment with its corresponding origin and routing said segment to its corresponding destination device;

if a portion of said data stream is not identified, means for sending said portion to a buffer;

means for processing the contents of said buffer to identify remaining segments with their corresponding origins and routing said segments to corresponding destination devices.

6. The apparatus of claim 5, wherein said processing means further comprises:

means for indicating a change in data type, said indicating means thereupon identifying a data segment having an origin;

means for matching said data segment to stored prototypes of data types;

if said matching identifies an origin of said data segment, means for routing said data segment to one or more destination devices corresponding to said origin;

if said matching does not identify an origin of said data segment, means for either discarding or storing said data segment.

7. The apparatus of claim 6, wherein said stored prototypes of data types are HMM, and wherein said matching of said data segment further comprises:

means for computing the best HMM path for said data segment for each HMM prototype;

means for comparing likelihood scores of best paths for all HMM prototypes; and

means for labeling said data segment with a data type that corresponds to the HMM having the highest likelihood score.

8. The apparatus of claim 6, wherein elements of said apparatus are related to OSI layers as follows:

said means for identifying data segments with known protocols are located in a physical link layer;

said data stream is represented as bit strings of “

1”

s and “

0”

s in a data link layer, said buffer for storage of these bit strings being located in a transport layer;

said means for segregating data types are located in a presentation layer;

said means for routing data segments to destination devices are located in an application layer.

9. The apparatus of claim 8, wherein said HMM based segregation system, said buffer, and a counter are embedded in a modem on a printed circuit card.

10. The apparatus of claim 6, wherein said indicating means further comprise:

means for associating a data type of first portions of said data stream with some of said data type prototypes;

means for computing a likelihood that a next data portion belongs to the data type of said first data portion;

wherein a change of data type between said first and said next data portions is indicated if the likelihood is below a precomputed threshold; and

wherein if said likelihood is above said precomputed threshold then said next data portion is associated with the data type of said first data portion, and the process is repeated with new data portions (until either data change is indicated or there are no new data portions).

11. The apparatus of claim 5, wherein said processing means further comprises:

means for classifying said buffer contents into a plurality of classes of data types, said classes being arranged in a hierarchy composed of a plurality of nodes and levels, the last level being said data types;

means for routing said buffer contents for processing at successive levels of said hierarchy, classification for any said level being done at a prior level;

means for segregating data types, said means being applied at a plurality of nodes in said hierarchy at the next level before said last level.

12. The apparatus of claim 11, wherein each said segregating means is further comprised of:

means for indicating a change in data type, said indicating means thereupon identifying a data segment having an origin;

means for matching said data segment to stored prototypes of data types;

if said matching identifies an origin of said data segment, means for routing said data segment to one or more destination devices corresponding to said origin;

if said matching does not identify an origin of said data segment, means for either discarding or storing said data segment.

13. The apparatus of claim 12, wherein said stored prototypes of data types are HMM, and wherein said matching of said data segment further comprises:

means for computing the best HMM path for said data segment for each HMM prototype;

means for comparing likelihood scores of best paths for all HMM prototypes; and

means for labeling said data segment with a data type that corresponds to the HMM having the highest likelihood score.

14. The apparatus of claim 13, wherein said HMM has the following features:

arc output labels are “

0”

s and “

1”

s;

each HMM is a concatenation of the same self loop and transition loop and models the arcs with a mixture of Gaussian distribution.

15. The apparatus of claim 14, wherein each HMM accepts as input a string of bits (“

0”

s and “

1”

s).

16. The apparatus of claim 13, wherein said prototypes correspond to the following data types:

voice, non-voice, speech, music, biometrics, textual, graphics, animated, telephone speech, and fax.

17. The apparatus of claim 11, wherein said hierarchy is of two levels, there being two classes of data types:

voice data and non-voice data.

18. The apparatus of claim 17, wherein voice data is segregated into data types including:

speech, music, telephone speech, and CD-ROM quality speech.

19. The apparatus of claim 17, wherein non-voice data is segregated into data types including:

textual, graphics, and fax.

20. The apparatus of claim 5, wherein said further identification is accomplished by a data classifier, which labels the identified segment with its corresponding origin and sends said segment to a router for routing to said segments corresponding destination device.

21. The apparatus of claim 1, wherein a user of the apparatus can send several different data types to said network in a single session without changing data type communication settings.

22. The apparatus of claim 1, wherein said means for processing and creating further comprise:

means for sampling allowable data types; and

means for building a data model using said sampling.

23. The apparatus of claim 22, wherein said building means uses Hidden Markov Models.

24. An apparatus for segregation of signals of different origin, comprising:

one or more computer systems at a receiving end, each having a central processing unit (CPU) and a memory, and each being connected to a network;

one or more communication devices that send or receive a plurality of data via said network, said data being of a plurality of origins and combined into a common data stream at said receiving end;

means, executed by said CPU, for evaluating portions of said data stream in sequence, and for generating data prototypes for each type of data processed at said receiving end;

means for storing said data prototypes in said memory.

25. The apparatus of claim 24, wherein data types for which said prototypes are created are the following:

voice, non-voice, textual, pictorial, biometrics, and graphics.

26. The apparatus of claim 25, where said data type prototypes are associated with data type HMM models.

27. The apparatus of claim 25, wherein said means for generating HMM prototypes further comprise:

means for forming an HMM topology;

means for training HMM output and transition arcs.