Method for processing and acquiring human voice data

Abstract

The invention discloses a method for processing and acquiring human voice data. The method comprises the following steps: acquiring a voice signal acquired by a collector of a mobile terminal; collecting speech data of the sampled data from a speech-based device; storing a general acoustic model and a personal acoustic model of the sampled data in a second memory, wherein the second memory is connected with a tone conversion unit for converting the tone of the sampled data into other required tones, the tones being selecting from a database, the database being stored in a second memory; when avoice recognition request is received from the sampled data, extracting a feature vector from the voice data of the sampled data; selecting any one of the general acoustic model and a personal acoustic model of the sampled data based on a cumulant of speech data of the sampled data; and recognizing a voice command by using the extracted feature vector and the selected acoustic model. The method is ingenious in design, reasonable in method, capable of reasonably processing human voice and suitable for application and popularization.

Description

Technical field [0001] The present invention relates to the technical field of methods for processing and acquiring human voice data, and in particular to a method for processing and acquiring human voice data. Background technique [0002] According to the conventional voice recognition method, voice recognition is performed using an acoustic model that has been stored in a voice recognition device in advance. The acoustic model is used to represent the attributes of the speaker's speech. For example, phonemes, diphones, triphones, pentaphones, syllables, and characters are used as the basic units of the acoustic model. If phonemes are used as the basic model of an acoustic model, since the number of acoustic models is reduced, context-sensitive acoustic models such as diphones, triphones, or pentaphones are widely used to reflect the synergy caused by changes between adjacent phonemes The phenomenon of coarticulation. Large amounts of data are needed to learn context-sensiti...

AI Technical Summary

Problems solved by technology

However, in this method, since the pitch of the speaker who actually uses the speech recognition function is often different from the pitch corresponding to the collected speech data, it is difficult to collect a large amount of speech data and ensure the speech recognition performance

Therefore, since an acoustic model is generally generated by learning voice data of an adult male, it is difficult to recognize voice commands of an adult female, an elderly person, or a child having a different voice pitch, and the recognized voice cannot be reused, let alone Carry out intelligent man-machine dialogue on the recognized voice, for this reason, the present invention proposes a method for processing and obtaining human voice data

Images