Method and device for voice recognition

• US 20050038652A1
• Filed: 12/21/2001
• Published: 02/17/2005
• Est. Priority Date: 12/21/2001
• Status: Active Grant

• Current Assignee: Telefonaktiebolaget LM Ericsson
• Original Assignee: Telefonaktiebolaget LM Ericsson
• Inventors: Dobler, Stefan
• Granted Patent: US 7,366,667 B2
• Time in Patent Office: Days
• US Class Current: 704/251
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 25/87 Detection of discrete points within a voice signal

Abstract

Method and device for the recognition of words and pauses in a voice signal. The words (W_i) spoken in a row and pauses (T_i) are thereby combined as to be appertaining to a word group as soon as one of the pauses (T_i) exceeds a limit value (TG). Stored references (R_j) are allocated to the voice signal of the word group, and an indication of the result of the allocation is effected after the limit value (TG) has been exceeded. To this end, parameters corresponding to the moments of the transitions between ranges with voice and non-voice are determined from the voice signal, and the limit value (TG) is then changed in dependence on said parameters.

Claims

1. Method for the recognition of words and pauses from a voice signal, wherein

the consecutively spoken words (W_i) and pauses (T_i) are combined as to be appertaining to a word group as soon as one of the pauses (T_i) exceeds a limit value (TG),

stored references (R_j) are allocated to the voice signal of the word group, and

an indication of the result of the allocation is effected after the limit value (TG) has been exceeded,

characterized in that

parameters corresponding to the moments of the transitions between ranges with voice and non-voice are determined from the voice signal, and

the limit value (TG) is changed in dependence on said parameters.

2. Method according to claim 1,

characterized in that

the moments of the transitions are the word beginning (WA_i) and the word end (WE_i) of the words (W_i) of the word group.

3. Method according to claim 1 and 2,

characterized in that

parameters of the references (R_j) are the word beginning (RA_j) and the word end (RE_j) of the stored references (R_j).

4. Method according to claim 3,

characterized in that

the limit value (TG) is increased, if at least for one word (W_i) from the word group the time interval between the word beginning (WA_i) and the word end (WE_i) is longer than the time interval between the word beginning (RA_j) and the word end (RE_j) of the allocated reference (R_j), and

the limit value (TG) is reduced, if at least for one word (W_i) from the word group the time interval between the word beginning (WA_i) and the word end (WE_i) is shorter than the time interval between the word beginning (RA_j) and the word end (RE_j) of the allocated reference (R_j).

5. Method according to claim 2,

characterized in that

the limit value (TG) is increased, if at least for one pause (T_i) from the word group the time interval between the word end (WE_i) of a word (W_i) and the word beginning (WA_i+1) of the subsequent word (W_i+1) is longer than a pause time interval (TP), and

the limit value (TG) is reduced, if at least for one pause (T_i) from the word group the time interval between the word end (WE_i) of a word (W_i) and the word beginning (WA_i+1) of the subsequent word (W_i+1) is shorter than a pause time interval (TP).

6. Method according to claim 5,

characterized in that

the pause time interval (TP) is a multiple of a reference pause model (R_p), wherein the reference pause model (R_p) is determined by the parameters word beginning (RA_j) and word end (RE_j) of a stored reference (R_j).

7. Method according to claim 1 to 6,

characterized in that

the moments of the transitions are the pause beginning (TA_i) and the pause end (TE_i) of the pauses (T_i) of the word group.

8. Method according to claim 7,

characterized in that

the limit value (TG) is increased, if at least for one pause (T_i) from the word group the time interval between the pause beginning (TA_i) and the pause end (TE_i) is longer than a pause time interval (TP), and

the limit value (TG) is reduced, if at least for one pause (T_i) from the word group the time interval between the pause beginning (TA_i) and the pause end (TE_i) is shorter than a pause time interval (TP).

9. Method for the recognition of words and pauses according to one of claims 1 to 8,

characterized in that

in a number of consecutive word groups the method is executed for each of the word groups.

10. Device for the recognition of words and pauses from a voice signal, wherein the words are combined as to be appertaining to a word group as soon as one of the pauses exceeds a limit value (TG), comprising a pattern comparator (120) provided for allocating at least one of the references stored in a memory unit (130) to the voice signal of the word group and the pattern comparator (120) being connectable with an input unit (200) for inputting the voice signal, and an output unit (300) for indicating the result of the allocation,

characterized in that

a control means (140) is provided to change the limit value (TG) in dependence on parameters of the voice signal, wherein the parameters correspond to the moments of the transitions between ranges with voice and non-voice.

11. Device according to claim 10,

characterized in that

pre-processing means (110) are provided to convert the words inputted as voice signal into an input unit (200) into a parametric description of the words for inputting the same into the pattern comparator (120).

12. Device according to claim 10 or 11,

characterized in that

the pattern comparator (120) allocates references to the voice signal by comparing the parameters (WA_i, WE_i) of the spoken words (W_i) with parameters (RA_j, RE_j) of the references (R_j), and

the control means (140) changes the limit value (TG) in dependence on the relation between the parameters of at least of one of the spoken words and the parameters of the allocated reference.

13. Device according to claims 10 to 12,

characterized in that

a detector (VAD) is provided to detect the parameters in the voice signal, which correspond to the moments of the transitions between ranges with voice and non-voice.

14. Device according to claim 13,

characterized in that

the control means (140) detects ranges as pauses from the parameters of the voice signal and changes the limit value (TG) in dependence on the relation between at least one of the pauses and one of the pause time intervals (TP) stored in the control means.

15. Portable terminal comprising a device according to one of claims 10 to 14.

16. Portable terminal according to claim 15, wherein the portable terminal is a mobile phone.

17. Central speech recognition device for a telecommunication network comprising a device according to one of claims 10 to 14.