中文摘要：

为提高机器学习性能,解决语音识别系统在噪声环境中识别率变差等问题,在分析支持向量机（SVM）模型抗噪性的基础上,提出一种基于生境共享机制的并行结构人工鱼群算法（PAFSA）优化SVM参数的方法。该算法对人工鱼群算法的循环主体进行改进,结合小生境技术的共享机制,在寻优的过程中维持样本个体的多样性,提高求解速度和解的精确性,并利用测试函数对该优化方法进行测试和比较,证明其有效性;用PAFSA对SVM中的惩罚因子C及高斯核参数γ进行优化,并将优选的参数用于一个非特定人、孤立词、中等词汇量的语音识别系统中。实验结果表明：当工作在不同信噪比和不同词汇量下,基于PAFSA-SVM模型语音识别率与基本AFSA-SVM模型识别率以及传统的HMM模型识别率相比均有不同程度提高。

英文摘要：

In order to improve the learning ability of the support vector machine （SVM）, and solve the problem that recognition rates of the speech recognition system become worse in the noisy environments, the noise immunity of SVM model was analyzed, and a parallel artificial fish swarm algorithm （PAFSA） was proposed based on niche technology to optimize the penalty parameter C and Gaussian kernel parameter y of SVM. The method improved the loop body of artificial fish swarm algorithm （AFSA）, combined with niche sharing mechanism in the optimization process to maintain the diversity of the sample of individuals, to improve the accuracy of the solution. By using some test functions, it is proved that its feasibility and effectiveness were improved, some optimized parameters were used in a non-specific persons, isolated words, and medium-vocabulary speech recognition system. The experimental results show that the speech recognition correct rates based on SVM using the PAFSA optimization parameters are better than those by the AFSA optimization parameters and traditional HMM model when it works under different SNRs and different words.