针对目前混淆网络生成算法速度与精度不能兼顾的不足,提出一种新的汉语音节混淆网络生成的方法.本算法采用类似轴对齐算法,对音节网格每次提取一条局部路径与参考路径对齐,根据每次对齐路径与参考路径长度不同,采用不同的策略生成混淆网络,并在生成混淆网络之后对其应用一种新的解码框架进行重打分.实验表明,该算法生成的混淆网络精度较高,时间复杂度优于轴对齐算法,且重打分后的识别率有显著提高. Aiming at the insufficiency of confusing the speed and accuracy of network generation algorithm, this paper proposes a new generation method of Chinese syllabification confusion network.The algorithm uses a similar axis alignment algorithm to extract a local path and a reference path at a time, According to the difference between the alignment path and the reference path length, a different strategy is used to generate the confusion network and a new decoding framework is applied to generate the confusion network after the confusion network is generated. The experiments show that the confusion network generated by this algorithm has higher accuracy , The time complexity is better than the axis alignment algorithm, and the recognition rate after heavy scoring is significantly improved.