Method of and device for segmenting an electric signal derived from an acoustic signal

• US 4,813,074 A
• Filed: 11/14/1986
• Issued: 03/14/1989
• Est. Priority Date: 11/29/1985
• Status: Expired due to Fees

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Marcus, Stephen M.
• Primary Examiner(s): Salce, Patrick R.
• Assistant Examiner(s): VOELTZ, EMANUEL T
• Application Number:
  US06/930,835
• Time in Patent Office: 851 Days
• Field of Search: 381/29-51, 364/513, 364/513.5
• US Class Current: 704/254
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection

Abstract

A method and device for segmenting an electric (acoustic) signal. A number of parameters, for example "log area" parameters, are derived as a function of time from the electric signal (block 2). For consecutive instants the signals of the parameters located within a time interval around each instant are taken up and transition functions (.0._i ) are derived therefrom for each instant (block 3). Subsequently similarity factors (c_ij) are determined each between two transition functions (.0._i and .0._j) associated with respective proximate instants (i and j) (block 4) and a replacing transition function (.0._ij^c) is derived (block 6) from the transition functions of at least one group of two transition functions for which the associated similarity factor is larger than a reference value V, the transition function (.0._ij^c) replace the aforementioned transition functions from which it is derived (block 7).

A combined transition function (.0._ij^c) replacing one of these transition functions is preferably calculated from the aforementioned transition functions, while the other of these transition functions are omitted. The group of transition functions for which the similarity factor is largest is preferably taken (block 5). The method is subsequently repeated (FIG. 1). Transition functions obtained are very suitable for use in the transmission or recognition of speech signals and for coding diphones.

Claims

1. A method of segmenting an electric signal derived from an acoustic signal into segments of the electric signal in which consecutively

a number of parameters is derived from the electric signal in the form of a number of signals as a function of time,

the signals of the parameters located within a time interval around each instant are taken up for consecutive instants,

a transition function .0. is determined for each instant from the signals of the parameters located within the time interval corresponding to said instant, and

information relating to the segments of the acoustic signal is derived from the transition functions associated with the instants,

characterized in that:

(a) similarity factors C_ij are determined, each between two transition functions (.0._i and .0._j) associated with respective proximate instants i and j, and

(b) a replacing transition function is derived from the transition functions of at least one group of two transition functions for which the associated similarity factor is larger than a reference value V, the said replacing transition function replacing said transition functions from which it is derived.

2. A method as claimed in claim 1, characterized in that the reference value V is a number between the largest similarity factor and the next to the largest similarity factor.

3. A method as claimed in claim 1 or 2, in which in step (b) the two transition functions of only one group are combined and a combined transition function replacing the said two transition functions is derived from said two transition functions, characterized in that (c) the combined transition function replaces one of the two transition functions from which the combined transition function is derived and the other of the two transition functions is omitted,

(d) in that subsequently similarity factors are determined between the combined transition function and the transition functions associated with instants in the proximity of the instant with which the combined transition function is associated, and in that the similarity factors associated with those combinations of two transition functions of which at least one is one of the two transition functions from which the combined transition function is derived, are omitted.

4. A method as claimed in claim 3, characterized in that (e) if there is subsequently at least one similarity factor which is larger than said reference value, the transition functions associated with such similarity factor are combined and a combined transition function replacing the aforementioned two transition functions is derived from the two transition functions, whereafter the method is continued with step (c),

(f) in that the reference value is decreased by a constant value if there are no similarity factors which are larger than the reference value,

(g) in that it is subsequently determined whether the reference value thus obtained is larger than a given constant,

(h) in that the method is continued with step (e) if the reference value is larger than the said constant, and

(i) in that the method is ended if the reference value is smaller than the said constant.

5. A method as claimed in claim 3, characterized in that it is determined whether the largest similarity factor is larger than a given constant,

(e) in that, if the largest similarity factor is larger than said constant, a combined transition function is derived from transition functions associated with this similarity factor, the combined transition function replacing the said two transition functions from which it is derived whereafter the method is continued with step (c), and

(f) in that the method is ended if the largest similarity factor is smaller than or equal to said constant.

6. A method as claimed in claim 1 or 2, in which in step (b) the two transition functions of only one group are combined and a combined transition function replacing said two transition functions is derived from these two transition functions, characterized in that in step (b) the combined transition function .0._ij^c is calculated in the following manner from the two transition functions .0._i and .0._j,

7. A method as claimed in claim 4 or 5, characterized in that in step (e) the combined transition function .0._ij^c is calculated from the two transition functions .0._i and .0._j in the following manner:

space="preserve" listing-type="equation">.0..sub.ij.sup.c =(n.sub.i .0..sub.i +n.sub.j .0..sub.j)/(n.sub.i +n.sub.j)

in which n_i and n_j indicate numbers of the original transition functions from which the transition functions .0._i and .0._j respectively are obtained by combining two transition functions in one or more previous combination steps.

8. A method as claimed in claim 7, characterized in that a number n_c which is equal to n_i +n_j is appended to the combined transition function, which number n_c indicates from how many original transition functions the combined transition function .0._ij^c has been obtained by combining two transition functions in one or more previous combination steps,

in that the number n_c replaces the number (n_i or n_j) associated with the transition function (.0._i or .0._j) which is replaced by the combined transition function, and in that the number (n_j or n_i), which is associated with the one of the two transition functions that is omitted, is also omitted.

9. A method as claimed in claim 1 or 2, characterized in that for calculating the similarity factor C_ij between two transition functions .0._i and .0._j, the transition functions are considered to be vectors in a multidimensional space and in that the similarity factor is equal to the reciprocal value of the euclidic distance between ends of said vectors.

10. A method as claimed in claim 4, characterized in that for calculating the similarity factor C_ij between two transition functions .0._i and .0._j the transition functions are considered to be vectors in a multidimensional space and in that the similarity factor is equal to the cosine of the angle between the two vectors.

11. A method as claimed in claim 10, characterized in that the said constant has a value of between 0.5 and 0.8.

12. A device for segmenting an electric signal derived from an acoustic signal comprising:

first means for deriving a number of parameters in the form of signals from the electric signal as a function of time,

second means for deriving a transition function for each instant from the signals of the parameters located within a time interval corresponding to said instant,

a first memory for storing the transition functions,

third means for calculating a similarity factor between two transition functions associated with respective proximate instants i and j,

a second memory for storing the similarity factors, and

fourth means for deriving a replacing or combined transition function from the transition functions of at least one group of transition functions for which the associated similarity factor is larger than a reference value.

13. A device as claimed in claim 12, further comprising a decision unit having an input coupled to an output of the second memory and an output coupled to an address input of the first memory.

14. A device as claimed in claim 12 or 13, characterized in that an output of the first memory is coupled to an input of the fourth means, and means coupling an output of the fourth means to an input of the first memory.

15. A device as claimed in claim 12 or 13 which further comprises a third memory for storing numbers n_i associated with transition functions .0._i, means coupling the output of the decision unit to an address input of the third memory, and means coupling an output of the third memory to another input of the fourth means.

16. A device as claimed in claim 15, wherein the fourth means are also adapted to derive a number n_c from numbers n_i and n_j and to apply said number to a second output, that is coupled to an input of the third memory.

17. A device as claimed in claim 14 further comprising:

a third memory for storing numbers n_i associated with transition functions .0._i, means coupling the output of the decision unit to an address input of the third memory, and means coupling an output of the third memory to another input of the fourth means.

18. A device as claimed in claim 17 wherein said fourth means derives a number n_c from numbers n_i and n_j and applies said number n_c to a second output that is coupled to an input of the third memory.