In this paper, a method of noise-robust speech emotion recognition under music noises is proposed by using a denoising autoencoder (DAE) and a support vector machine (SVM). The proposed method first trains a DAE by using emotional speech signals corrupted by music noises. Then, the output values from a middle layer of the DAE are used as speech features. Next, an SVM is trained to classify emotions using the DAE features. The performance of the proposed method is compared with that of a conventional SVM classifier. Consequently, it is shown that the proposed method relatively improves the overall emotion recognition rate by 9.76% under music noise conditions, compared to the conventional method.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
