Method and device for the processing of speech information

• US 7,177,800 B2
• Filed: 11/02/2001
• Issued: 02/13/2007
• Est. Priority Date: 11/03/2000
• Status: Expired due to Fees

• Current Assignee: Digital Design GmbH
• Original Assignee: Digital Design GmbH
• Inventors: Wallers, Joseph
• Primary Examiner(s): Chawan; Vijay B.
• Application Number:
  US09/985,430
• Publication Number:
  US 20020082841A1
• Time in Patent Office: 1,929 Days
• Field of Search: 704/201, 704/270, 704/270.1, 704/235, 704/260, 704/2, 704/4, 704/1, 704/9, 707 1- 5, 369/44.28, 369/53.24
• US Class Current: 704/201
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering

Abstract

A method and a device for processing of speech information uses the input and/or storage and/or the acoustical reproduction and/or for the transmission of speed and data information to other devices for local storage and/or reproduction, as well as searching for one or several speech segments in the stored speech information. A recording and search/reproduction of speech information is ensured without manual designation and classification and without the requirement of a vocabulary. Spoken words and/or correlated sentences (memorandums) are digitally recorded as speech signals in a memory which, in a partial scope of at least one word, are again spoken for search purposes and are compared in a device with the recordings and subsequently evaluated, from which a score between the two speech patterns is determined and the memorandum with the smallest score is issued/reproduced acoustically.

Claims

1. A method for the processing of speech information for the realisation of a speech based memory using at least one of means for input, means for storage, means for acoustical reproduction, means for transmission of speech and data information to other devices and means for searching for one or several speech segments in speech information stored in the means for storage, the method comprising the steps of:

recording digitally at least one spoken voice memorandum as speech signals in the means for storage;

transforming speech signals into at least one representation suitable for pattern matching;

recording at least one search word spoken for search purposes;

comparing the at least one recorded search word with the digital recordings of the at least one voice memorandum with the means for searching by way of pattern matching;

determining a score between the speech patterns of the search word and the digital recordings of the at least one voice memorandum on a basis of their parameters; and

sorting at least a part of the at least one voice memorandum according to the score.

2. The method according to claim 1, wherein speech signals are recorded in a compressed manner.

3. The method according to claim 1 or 2, wherein non-compressed signals are stored in addition to compressed speech information.

4. The method according to claim 1, wherein supplementary speech signals are stored associated with at least one of the recorded voice memorandums.

5. The method according to claim 1, wherein during a search operation, speech patterns are compared with the same data records, which are also used for reproduction purposes.

6. The method according to claim 1, wherein reactions of the user as to results of a search operation are recorded.

7. The method according to claim 1, wherein during search operations, speech patterns of former search commands are searched through, taking into consideration recorded evaluations and indicators.

8. The method according to claim 1, wherein during the reproduction of a found item, the search operation is continued.

9. The method according to claim 1, wherein search algorithms and parameters are optimized on the basis of the recorded patterns and evaluations.

10. The method according to claim 1, wherein storage is performed in an encrypted manner and an access protection is installed.

11. The method according to claim 1, wherein for input of speech signals, a microphone, telephone dictating machine or voice box are used.

12. The method according to claim 1, wherein for reproduction, a headphone or telephone are used.

13. The method according to claim 1, wherein a short-time storage is performed in a mobile phone, and a long-time storage is performed on a server, where data are transmitted between short-time storage and long-time storage periodically r with access to the long-time storage.

14. The method according to claim 1, wherein an index is established in which individual speech patterns are separately filed, the individual speech patterns being provided with at lest one of indicators to recorded voice memorandums and coefficients of coincidences (scores).

15. The method according to claim 14, wherein the index pattern is established by a user by speaking individual words.

16. The method according to claim 14 or 15, wherein the index is established offline.

17. The method according to claim 14, wherein an offline optimization or index generation is carried out when the mobile phone is connected up to the power supply.

18. The method according to claim 14, wherein an offline optimization is carried out on a high-performance computer.

19. The method according to claim 1, wherein search operation is performed with a time indication.

20. The method according to claim 1, wherein data files are recorded with link to at least one of the recorded voice memorandums.

21. A device for processing speech information for realization of a speech based memory comprising:

input means for speech signals;

storage means for storing digitally at least one voice memorandum;

means for transforming speech signals into at least one representation suitable for pattern matching;

means for searching one or several speech segments in speech information, where the means for searching are arranged

to compare at least one recorded search word with the digital recordings of the at least one voice memorandum with the means for searching by way of pattern matching,

to determine a score between the speech patterns of the search word and the digital recordings of the at least one voice memorandum on the basis of their parameters, and

to sort at least a part of the at least one voice memorandum according to the score; and

output means for outputting the results of at least one of search and sorting.

22. The device according to claim 21, wherein the device comprises a telecommunication device with a memory card.

23. The device according to claim 22, wherein the device comprises a data processing unit equipped with a special additional software.

24. The device according to the claim 21, wherein the device comprises a computer with speech input means and speech output means, where the computer is equipped with special software.

25. The device according to the claim 21, wherein the device comprises a telecommunication device connected by way of a network to a computer or a special server.