成都工业学院学报
成都工業學院學報
성도공업학원학보
Journal of Chengdu Technological University
2015年
2期
32-34
,共3页
CFCC-PCA%说话人辨识%支持向量机%云服务器
CFCC-PCA%說話人辨識%支持嚮量機%雲服務器
CFCC-PCA%설화인변식%지지향량궤%운복무기
CFCC-PCA%speaker identification%Support Vector Machine( SVM)%cloud server
针对说话人训练和识别时间长、噪音环境下识别率低的问题,提出一种CFCC-PCA特征参数的说话人辨识方法。首先提取具有听觉特性的CFCC特征参数,然后对其进行PCA变换,找出具有分辨能力的参数,最后再用这些参数在云服务器中训练和识别说话人。实验表明:该方法可以提高说话人辨识的鲁棒性和识别率,云服务可提高系统实时性。
針對說話人訓練和識彆時間長、譟音環境下識彆率低的問題,提齣一種CFCC-PCA特徵參數的說話人辨識方法。首先提取具有聽覺特性的CFCC特徵參數,然後對其進行PCA變換,找齣具有分辨能力的參數,最後再用這些參數在雲服務器中訓練和識彆說話人。實驗錶明:該方法可以提高說話人辨識的魯棒性和識彆率,雲服務可提高繫統實時性。
침대설화인훈련화식별시간장、조음배경하식별솔저적문제,제출일충CFCC-PCA특정삼수적설화인변식방법。수선제취구유은각특성적CFCC특정삼수,연후대기진행PCA변환,조출구유분변능력적삼수,최후재용저사삼수재운복무기중훈련화식별설화인。실험표명:해방법가이제고설화인변식적로봉성화식별솔,운복무가제고계통실시성。
Training speaker system and speaker identification need a long time, and in the noise environment, the recognition rate is very low, A CFCC-PCA characteristic parameter method is proposed.Firstly, the acoustic characteristics of CFCC characteristic parameters are extracted.Then, CFCC-PCA parameters are extracted by PCA transformation of CFCC characteristic parameters.Finally the speaker models are trained and recognized in cloud.Experiments show that the CFCC-PCA characteristic parameters can improve the robustness and recognition rate of the speaker, the cloud services with efficient processing ability to improve system real-time performance.