论文部分内容阅读
针对最小二乘支持向量机在语音数据大样本输入下的局限性,提出了一种基于GMM特征变换和模糊LS-SVM的学习算法,结合高斯混合模型在拟合数据分布方面和最小二乘支持向量机在分类辨别方面的突出优势,有效地提取说话者特征信息,压缩了数据,解决了大样本数据输入下最小二乘支持向量机的训练速度和测试精度问题,同时在LS-SVM系统中引入模糊隶属度函数,很好地解决了不可分数据的输出.理论研究和实验表明,所提方法能充分地利用训练数据,使得系统在具有更好辨别能力的同时提高了鲁棒性和识别率.
Aiming at the limitation of least squares support vector machine (SVM) under the large input of speech data, a learning algorithm based on GMM feature transform and fuzzy LS-SVM is proposed. Combining Gaussian mixture model with fitting data distribution and least square support In the LS-SVM system, vector machine has outstanding advantages in classification discrimination, effectively extracting speaker characteristic information, compressing the data, and solving the training speed and testing accuracy of least square support vector machine with large sample data input. The fuzzy membership function is introduced to solve the problem of indivisible data output theoretically and experimentally.The results show that the proposed method can make full use of the training data and improve the robustness and recognition rate .