VOICE SEARCH DEVICE, VOICE SEARCH METHOD, AND NON-TRANSITORY RECORDING MEDIUM

• US 20150255059A1
• Filed: 01/23/2015
• Published: 09/10/2015
• Est. Priority Date: 03/05/2014
• Status: Active Grant

• Current Assignee: Casio Computer Company Limited
• Original Assignee: Casio Computer Company Limited
• Inventors: IDE, Hiroyasu
• Granted Patent: US 9,437,187 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/632 Query formulation
• CPC Class Codes: G06F 16/683 using metadata automatically derived from the content
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units
• CPC Class Codes: G10L 2015/081 Search algorithms, e.g. Baum-Welch or Viterbi
• CPC Class Codes: G10L 25/54 for retrieval
• CPC Class Codes: G10L 25/87 Detection of discrete points within a voice signal

Abstract

A search string acquiring unit acquires a search string. A converting unit converts the search string into a phoneme sequence. A time length deriving unit derives the spoken time length of the voice corresponding to the search string. A zone designating unit designates a likelihood acquisition zone in a target voice signal. A likelihood acquiring device acquires a likelihood indicating how likely the likelihood acquisition interval is an interval in which voice corresponding to the search string is spoken. A repeating unit changes the likelihood acquisition zone designated by the zone designating unit, and repeats the process of the zone designating unit and the likelihood acquiring device. An identifying unit identifies, from the target voice signal, estimated intervals for which the voice corresponding to the search string is estimated to be spoken, on the basis of the likelihoods acquired for each of the likelihood acquisition zones.

Claims

1. A voice search device comprising:

a search string acquirer acquiring a search string;

a converter converting the search string acquired by the search string acquirer into a phoneme sequence;

a time length deriver acquiring a duration of each phoneme included in the phoneme sequence converted by the converter, and deriving a spoken time length of voice corresponding to the search string based on the acquired durations;

a zone designator designating a likelihood acquisition zone that is a zone of the time length derived by the time length deriver in a target voice signal;

a likelihood acquirer acquiring a likelihood indicating how likely the likelihood acquisition zone designated by the zone designator is a zone in which voice corresponding to the search string is spoken;

a repeater changing the likelihood acquisition zone designated by the zone designator, and repeating a process of the zone designator and the likelihood acquirer; and

an identifier identifying, on the basis of the likelihood acquired by the likelihood acquirer for each likelihood acquisition zone designated by the zone designator, from the target voice signal an estimated zone for which the voice corresponding to the search string is estimated to be spoken.

2. The voice search device according to claim 1, further comprising:

an output probability acquirer acquiring, for each frame, an output probability of a feature quantity of the target voice signal being output from each phoneme included in the phoneme sequence;

wherein the likelihood acquirer acquires the likelihood of the likelihood acquisition zone based on the output probability acquired by the output probability acquirer in the likelihood acquisition zone designated by the zone designator.

3. The voice search device according to claim 2, further comprising:

a feature quantity calculator calculating, for each frame, a feature quantity of the target voice signal in the likelihood acquisition zone designated by the zone designator;

wherein the output probability acquirer acquires the output probability based on the feature quantity calculated by the feature quantity calculator.

4. The voice search device according to claim 2, further comprising:

an output probability storage storing each phoneme of an acoustic model in association with the output probability of the feature quantity of the target voice signal being output from the each phoneme, for every frame included in the target voice signal;

wherein, after the converter converts the search string into the phoneme sequence, the output probability acquirer acquires, from among output probabilities stored in the output probability storage, an output probability stored in association with each phoneme included in the phoneme sequence.

5. The voice search device according to claim 2, further comprising:

a replacer replacing the output probability acquired for each frame by the output probability acquirer with an output probability of maximum value from among a plurality of output probabilities acquired in a plurality of consecutive frames that includes the output probability;

wherein the likelihood acquirer acquires the likelihood of the likelihood acquisition zone designated by the zone designator based on an output probability replaced by the replacer.

6. The voice search device according to claim 2,

wherein the converter converts the search string into the phoneme sequence by sorting phonemes of an acoustic model that does not depend on adjacent phonemes, and

the likelihood acquirer acquires the likelihood of the likelihood acquisition zone designated by the zone designator based on the phoneme sequence,

the voice search device further comprising:

a selector selecting a plurality of candidates of the estimated zone from among likelihood acquisition zones designated by the zone designator, on the basis of likelihoods acquired by the likelihood acquirer;

a second converter converting the search string into a second phoneme sequence by sorting phonemes of a second acoustic model that depends on adjacent phonemes; and

a second likelihood acquirer acquiring, on the basis of the second phoneme sequence, a second likelihood indicating how likely a zone selected as the candidate of the estimated zone by the selector is a zone in which the voice corresponding to the search string is spoken, for each of the plurality of candidates selected by the selector;

wherein the identifier identifies the estimated zone from among the plurality of candidates selected by the selector on the basis of the second likelihood acquired by the second likelihood acquirer.

7. The voice search device according to claim 6, wherein

the selector selects a plurality of candidates of the estimated zone by selecting likelihood acquisition zone of maximum likelihood one by one from among likelihood acquisition zones included in a zone of a predetermined selection time length, for every predetermined selection time length.

8. The voice search device according to claim 1, wherein

the time length deriver derives a plurality of mutually different time lengths as a spoken time length of voice corresponding to one search string acquired by the search string acquirer,

the zone designator designates a likelihood acquisition zone that is a zone of a time length derived by the time length deriver in the target voice signal for each of the plurality of time lengths,

the likelihood acquirer acquires the likelihood of a likelihood acquisition zone designated by the zone designator for each of the plurality of time lengths, and

the identifier identifies, on the basis of the likelihood acquired by the likelihood acquirer, the estimated zone from among likelihood acquisition zones designated by the zone designator for each of the plurality of time lengths.

9. A voice search method comprising:

a search string acquiring step that acquires a search string;

a converting step that converts the search string acquired by the search string acquiring step into a phoneme sequence;

a time length deriving step that acquires a duration of each phoneme included in the phoneme sequence converted by the converting step, and derives a spoken time length of voice corresponding to the search string based on the acquired durations;

a zone designating step that designates a likelihood acquisition zone that is a zone of the time length derived by the time length deriving step in a target voice signal;

a likelihood acquiring step that acquires a likelihood indicating how likely the likelihood acquisition zone designated by the zone designating step is a zone in which voice corresponding to the search string is spoken;

a repeating step that changes the likelihood acquisition zone designated by the zone designating step, and repeats a process of the zone designating step and the likelihood acquiring step; and

an identifying step that identifies, on the basis of the likelihood acquired by the likelihood acquiring step for each likelihood acquisition zone designated by the zone designating step, from the target voice signal an estimated zone for which the voice corresponding to the search string is estimated to be spoken.

10. The voice search method according to claim 9, further comprising:

an output probability acquiring step that acquires, for each frame, an output probability of a feature quantity of the target voice signal being output from each phoneme included in the phoneme sequence;

wherein, in the likelihood acquiring step, the likelihood of the likelihood acquisition zone is acquired based on the output probability acquired by the output probability acquirer in the likelihood acquisition zone designated by the zone designator.

11. The voice search method according to claim 10, further comprising:

a feature quantity calculating step that calculates, for each frame, a feature quantity of the target voice signal in a likelihood acquisition zone designated in the zone designating step;

wherein, in the output probability acquiring step, the output probability is acquired based on the feature quantity calculated by the feature quantity calculating step.

12. The voice search method according to claim 10, further comprising:

an output probability storing step that stores each phoneme of an acoustic model in association with the output probability of the feature quantity of the target voice signal being output from the each phoneme, for every frame included in the target voice signal;

wherein, in the output probability acquiring step, after the converting step converts the search string into the phoneme sequence, an output probability stored in association with each phoneme included in the phoneme sequence is acquired from among output probabilities stored in the output probability storing step.

13. The voice search method according to claim 10, further comprising:

a replacing step that replaces the output probability acquired for each frame in the output probability acquiring step with an output probability of maximum value from among a plurality of output probabilities acquired in a plurality of consecutive frames that includes the output probability;

wherein, in the likelihood acquiring step, the likelihood of the likelihood acquisition zone designated by the zone designating step based on an output probability replaced by the replacing step is acquired.

14. The voice search method according to claim 10,

wherein, in the converting step, the search string is converted into the phoneme sequence by sorting phonemes of an acoustic model that does not depend on adjacent phonemes, and

in the likelihood acquiring step, the likelihood of a likelihood acquisition zone designated by the zone designating step is acquired based on the phoneme sequence,

the voice search method further comprising:

a selecting step that selects a plurality of candidates of the estimated zone from among likelihood acquisition zones designated by the zone designating step, on the basis of likelihoods acquired in the likelihood acquiring step;

a second converting step that converts the search string into a second phoneme sequence by sorting phonemes of a second acoustic model that depends on adjacent phonemes; and

a second likelihood acquiring step that acquires, on the basis of the second phoneme sequence, a second likelihood indicating how likely a zone selected as the candidate of the estimated zone by the selecting step is a zone in which the voice corresponding to the search string is spoken, for each of the plurality of candidates selected in the selecting step;

wherein the identifying step identifies the estimated zone from among the plurality of candidates selected in the selecting step on the basis of the second likelihood acquired by the second likelihood acquiring step.

15. The voice search method according to claim 14, wherein

in the selecting step, a plurality of candidates of the estimated zone are selected by selecting likelihood acquisition zone of maximum likelihood one by one from among likelihood acquisition zones included in a zone of a predetermined selection time length, for every predetermined selection time length.

16. The voice search method according to claim 9, wherein

in the time length deriving step, a plurality of mutually different time lengths are derived as a spoken time length of voice corresponding to one search string acquired by the search string acquiring step,

in the zone designating step, a likelihood acquisition zone that is a zone of a time length derived by the time length deriving step in the target voice signal is designated for each of the plurality of time lengths,

in the likelihood acquiring step, the likelihood of a likelihood acquisition zone designated by the zone designating step for each of the plurality of time lengths is acquired, and

in the identifying step, on the basis of the likelihood acquired by the likelihood acquiring step, the estimated zone is identified from among likelihood acquisition zones designated in the zone specifying step for each of the plurality of time lengths.

17. A non-transitory recording medium storing a program causing a computer to function as:

a search string acquirer acquiring a search string;

a converter converting the search string acquired by the search string acquirer into a phoneme sequence;

a time length deriver acquiring a duration of each phoneme included in the phoneme sequence converted by the converter, and deriving a spoken time length of voice corresponding to the search string based on the acquired durations;

a zone designator designating a likelihood acquisition zone that is a zone of the time length derived by the time length deriver in a target voice signal;

a likelihood acquirer acquiring a likelihood indicating how likely the likelihood acquisition zone designated by the zone designator is a zone in which voice corresponding to the search string is spoken;

a repeater changing the likelihood acquisition zone designated by the zone designator, and repeating a process of the zone designator and the likelihood acquirer; and

an identifier identifying, on the basis of the likelihood acquired by the likelihood acquirer for each likelihood acquisition zone designated by the zone designator, from the target voice signal an estimated zone for which the voice corresponding to the search string is estimated to be spoken.