本文设计了一种混合神经网络矢量量化编码方法,利用Kohonen网络的自组织聚类功能设计矢量量化器码书,实现矢量量化,用三层BP网络完成码字的信道符号编码.该神经网络矢量量化编码器能够并行处理输入矢量,速度快,效率高,适用于语音和图像数据压缩. In this paper, a hybrid neural network vector quantization coding method is designed. The vector quantizer codebook is designed by using the self-organizing clustering function of Kohonen network to realize vector quantization. The three-layer BP network is used to code the channel symbol of the codeword. Quantization Encoder can process input vector in parallel, fast, high efficiency, suitable for voice and image data compression.