Method of determining reference values

• US 5,937,382 A
• Filed: 05/02/1996
• Issued: 08/10/1999
• Est. Priority Date: 05/05/1995
• Status: Expired due to Fees

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Hab-Umbach, Reinhold
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Opsasnick, Michael N.
• Application Number:
  US08/642,015
• Time in Patent Office: 1,195 Days
• Field of Search: 395/2.6-2.66, 704/251-257
• US Class Current: 704/251
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 2015/0635 updating or merging of old and new templates; Mean values; Weighting

Abstract

Reference values are required for deriving word sequences from an unknown speech signal. These reference values are determined from a known test speech signal and represent multi-dimensional mean value vectors which are derived from the characteristic vectors formed during the processing of the test speech signal. As the processing of the test speech signal progresses mean value vectors are continuously split into two new mean value vectors, again. In order to minimize the total number of mean value vectors required for the description of the individual acoustic states of a speech signal, it is proposed to link individual mean value vectors, irrespective of whether or not they are associated with different acoustic states, provided that they neighbor one another closely enough. Notably if the steps of linking and splitting are performed a number of times, preferably in an alternating fashion, there is obtained a set of mean value vectors which approximate the acoustic states comparatively accurately. The mean value vectors, being associated with an acoustic state with each time the corresponding weight, constitute the reference value for the relevant acoustic state.

Claims

1. A method of determining reference signals associated with acoustic states and generated from a test speech signal during a test phase, the method comprising the steps of:

a) deriving, at regular time intervals, multi-dimensional characteristic vectors from the test speech signal and storing the derived characteristic vectors,

b) determining a mean value vector from neighboring characteristic vectors, each mean value vector representing the center of gravity of the neighboring characteristic vectors,

c) splitting a mean value vector into two new mean value vectors whenever a threshold condition for splitting the mean value vector is satisfied,

d) determining whether a threshold condition for linking mean value vectors is satisfied, the threshold condition for linking mean value vectors being a function of the set of all mean value vectors,

e) where the threshold condition for linking is satisfied, linking pairs of arbitrary neighboring mean value vectors that satisfy threshold conditions of proximity, independent of an acoustic state associated with the mean value vectors, by replacing the pairs with a single new mean value vector, such linking performed in direct succession for pairs of mean value vectors satisfying the threshold conditions of proximity, and

f) assigning mean value vectors to reference signals.

2. A method as claimed in claim 1, wherein the linking of the two arbitrary neighboring mean value vectors is performed repeatedly, notably each time after formation of two mean value vectors instead of a single mean value vector.

3. A method as claimed in claim 1, wherein;

more than one mean value vector is associated with at least a part of the acoustic states;

at least some mean value vectors are associated with two or more acoustic states; and

after the linking of two neighboring mean value vectors, into a new mean value vector the assignment of the new mean value vector to the acoustic states is determined anew.

4. A method as claimed in claim 2, wherein:

more than one mean value vector is associated with at least a part of the acoustic states;

at least some mean value vectors are associated with two or more acoustic stages; and

after the linking of two neighboring mean value vectors into a new mean value vector, the assignment thereof to the acoustic states is determined anew.