In order to overcome the limitation of speech visualization. This paper proposed a novel speech visualization method for Chinese vowel sound based on neural network. It created readable patterns by integrating different speech features into a single picture. It used wavelet neural network to map location information and color information. Because the wavelet neural network has the advantages of structure designability, convergence precision controllability and rapid convergence, that effectively improve the correct rate of Chinese vowel sound encoding. The image was divided into 12 different color display areas, the speech for each display area have similar pronunciation characteristics and the same pronunciation articulation. That make full use of the advantages of deaf people of visual identification ability and visual memory ability for color. Compared with the existing method , this method has good robustness and understandability.%为了克服现有语音可视化方法的局限性,该文提出了一种基于神经网络的汉语声韵母可视化方法,通过集成不同的语音特征进入一幅图像中为聋哑人创造了语音信号的可读模式.采用小波神经网络来进行位置信息映射和颜色信息获取,由于小波神经网络具有结构可设计性、收敛精度可控性和收敛速度快的优点,有效地提高了汉语声韵母的正确编码率.而且将图像分为12个不同颜色的显示区域,每个显示区域内的音具有相似的发音特点和相同的发音部位,这就更好地利用了聋哑人对色彩刺激的视觉记忆能力较强的优点.与现有方法相比,具有很好的鲁棒性和易懂性.
展开▼
