A method and device for distinguishing between machine voice and natural voice
A speech and machine technology, applied in speech analysis, speech recognition, instruments, etc., can solve complex or personalized problems that cannot be solved
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0067] In this embodiment, the morpheme segmentation of the collected speech to obtain the morpheme speech includes:
[0068] Perform speech recognition on the collected speech, and obtain the pronunciation interval between syllables of the collected speech;
[0069] One or more syllables whose pronunciation interval in the recognized speech is smaller than the preset interval value is regarded as a morpheme speech.
[0070] The syllables are defined according to the characteristics of the language. For example, for Chinese, the pronunciation of a Chinese character generally corresponds to one syllable, and the syllable usually includes initial consonants, finals and tones. Word pronunciation can correspond to one or more syllables.
[0071] The size of the preset interval value only needs to ensure that each syllable can be distinguished.
[0072] In one embodiment, the preset interval value can be adjusted according to the speech rate of the speech producer, and the preset...
Embodiment 2
[0075] In this embodiment, the acquisition of the number of morpheme sounds belonging to the same type of morpheme sounds from the segmented morpheme sounds includes:
[0076] Obtain the sound feature value of the morpheme speech obtained by segmentation;
[0077] Attribute the morpheme sounds with close sound feature values to the same morpheme sounds;
[0078] The morpheme voices belonging to the same morpheme voice are counted.
[0079] In one embodiment, the sound feature values include feature values obtained based on syllables.
[0080] In an embodiment, the same morpheme voices can be classified into the same morpheme voice set, and the number of morpheme voices belonging to the same morpheme voice can be obtained by counting the morpheme voices in the set.
[0081] In one embodiment, if the quantity in the same morpheme voice set exceeds the third preset value, the morpheme voices in the set can be further classified according to semantics, and the morpheme voi...
Embodiment 3
[0084] In this embodiment, the comparison of the number of morpheme sounds corresponding to the same type of morpheme sounds includes:
[0085] Obtaining the pronunciation length and / or intonation of morpheme sounds belonging to the same morpheme sounds from the collected speech;
[0086] Compare the pronunciation length and / or intonation of morpheme sounds that belong to the same type of morpheme sounds.
[0087] In one embodiment, the method further includes obtaining, from the collected speech, the pronunciation intervals of the morpheme speeches belonging to the same kind of morpheme speeches. The pronunciation interval can reflect the speech habit of the speech producer, especially a real person.
[0088] The technical solution for distinguishing between machine speech and natural speech proposed by the embodiments of the present invention helps to capture the voices of different speech producers (especially real people) by acquiring the pronunciation length and / or inton...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More