Speech recognition based mouth shape animation implementing method

Abstract

The invention provides a speech recognition based mouth shape animation implementing method and device. The device comprises an acquisition unit, a signal processing unit, an animation synthesizing unit and a display unit; the acquisition unit is configured to acquire an analog signal of original sound data; the signal processing unit is configured to analyze sound data to obtain vowel of the sound data; the animation synthesizing unit is configured to call a skeletal animation according to the vowel of the sound data, so as to allow the skeletal animation to make a mouth shape motion matchingthe vowel, and fuse and play the mouth shape motion and the existing animation to obtain the final mouth shape animation; and the display unit is configured to display the mouth shape animation generated by the animation synthesizing unit.

Description

technical field [0001] The invention relates to the technical field of interlingual recognition in the communication field, in particular to a method and device for implementing lip animation based on speech recognition. Background technique [0002] With the development of entertainment industries such as games and animation, in various game applications, animation videos or computer simulation applications, there are usually some dialogue scenes in which image display needs to be coordinated with audio. In these dialogue scenes, the scene characters take turns delivering dialogue. For example, a game application usually involves a game plot dialogue scene, and in the game plot dialogue scene, game characters will have dialogues in turn. It can be seen that in the dialogue scene, it is not only necessary to play the dialogue sound of the scene character, but also to present the scene character's mouth shape that matches the dialogue audio, that is, when the scene character...

AI Technical Summary

Problems solved by technology

But this solution needs to store a large number of mouth pictures, and the storage capacity needs to be very large

[0004] At present, there are also some methods that segment the sound time-domain waveform and compare the audio feature library of the existing vocabulary to identify the corresponding words. However, this recognition method cannot accurately give the time point of the appearance of a single word. There will also be different recognition results

Because the sound is an analog signal, the time-domain waveform of the obtained sound only represents the relationship between the sound pressure and the time change, and cannot represent the characteristics of the sound well, so the recognition accuracy is not high, and the matching degree of the final lip animation is also low. not too good

Images