基于OpenCV的口唇检测研究

Research on Lip Detection Based on OpenCV

唇读技术在提高噪音环境中语音识别系统的识别率和帮助有听力障碍的人更好的理解语言方面起到了一定的辅助作用。由于说话过程中口型变化较大,直接提取口型区域较困难;而在发音过程中鼻子形状不会发生较大变化,且OpenCV自带的人脸检测器可以很好的检测出人脸。提出了一种利用嘴唇与人脸和鼻子的相对位置关系进行检测嘴唇区域的方法。实验证明,该方法可以比较快速,精确地提取出嘴唇区域,从而利于精确的进行口唇分割。

Lip-reading systems play an important role to improve the recognition rate of the speech recognition systems on the noise environment and to help the disabled in hearing communicate with others.The shape of lip has obvious changes when somebody is pronouncing,so it is difficult to directly detect the lip region;on the contrary,the nose shape has not obvious changes,and the face detection in OpenCV software has achieved considerable effect on detection face,so a method of detection lip is presented which adopts the relative position of lip against to face and nose to detect the lip region.The experiments show that this method can quickly and efficiently extract the lip region,and help to improve the accuracy of lip segmentation.