论文部分内容阅读
针对支持向量机(Support Vector Machine,SVM)的参数优化问题,提出了一种改进的混合蛙跳算法(Improved Shuffled Frog Leaping Algorithm,Im-SFLA),提高了其在实用语音情感识别中的学习能力。首先,我们在SFLA中引入了模拟退火(Simulated Annealing,SA)、免疫接种(Immune Vaccination,IV)、高斯变异和混沌扰动算子,平衡了搜索的高效性和种群的多样性;第二,利用Im-SFLA优化SVM的参数,提出了一种Im-SFLA-SVM方法;第三,分析了烦躁等实用语音情感的声学特征,重点分析了基音、短时能量、共振峰和混沌特征随情感类别的变化特性,构建出144维的情感特征向量并采用LDA降维到4维;最后,在实用语音情感数据库上测试了算法性能,将提出的算法与混合蛙跳算法(Shuffled Frog Leaping Algorithm,SFLA)优化SVM参数的方法(SFLA-SVM方法)、粒子群优化(Particle Swarm Optimization,PSO)算法优化SVM参数的方法(PSO-SVM方法)、基本SVM方法、高斯混合模型(Gaussian Mixture Model,GMM)方法和反向传播(Back Propagation,BP)神经网络法等进行对比。实验结果表明,采用Im-SFLA-SVM方法的平均识别率达到77.8%,分别高于SFLA-SVM方法、PSO-SVM方法、SVM方法、GMM方法和BP神经网络法各1.7%,2.7%,3.4%,4.7%,7.8%,并且对于烦躁这种实用情感的识别率提高效果最为明显,从而证实了Im-SFLA是一种有效的SVM参数选择方法,并且Im-SFLA-SVM方法能显著提升实用语音情感的识别率。
In order to solve the problem of parameter optimization in Support Vector Machine (SVM), an improved improved Frog Leaping Algorithm (Im-SFLA) is proposed to improve its learning ability in practical speech emotion recognition . First of all, we introduced Simulated Annealing (SA), Immune Vaccination (IV), Gaussian mutation and chaotic perturbation operator into SFLA to balance the search efficiency and population diversity. Secondly, SF-SVA, Im-SFLA proposed an Im-SFLA-SVM method. Thirdly, the acoustic characteristics of practical speech emotion such as irritability were analyzed. The characteristics of pitch, short-time energy, formant and chaos were analyzed emphatically. And the dimensionality of 144 dimensions is constructed and reduced to 4 dimensions using LDA. Finally, the performance of the algorithm is tested on a practical speech emotion database. The proposed algorithm is compared with the Shuffled Frog Leaping Algorithm (SFLA (SFLA-SVM method), Particle Swarm Optimization (PSO) method for optimizing SVM parameters (PSO-SVM method), basic SVM method, Gaussian Mixture Model (GMM) Methods and Back Propagation (BP) neural network method for comparison. The experimental results show that the average recognition rate of Im-SFLA-SVM method is 77.8%, which is higher than SFLA-SVM method, PSO-SVM method, SVM method, GMM method and BP neural network method respectively by 1.7%, 2.7%, 3.4 %, 4.7% and 7.8%, respectively, and the most obvious effect is to improve the recognition rate of irritability. This proves that Im-SFLA is an effective SVM parameter selection method and that the Im-SFLA-SVM method can significantly improve the practicality Speech emotion recognition rate.