Kernel PCA and de-noising in feature spaces
Proceedings of the 1998 conference on Advances in neural information processing systems II
Evaluation of Objective Quality Measures for Speech Enhancement
IEEE Transactions on Audio, Speech, and Language Processing
The pre-image problem in kernel methods
IEEE Transactions on Neural Networks
Hi-index | 0.00 |
In this paper, we use kernel principal component analysis (kPCA) for speech enhancement. To synthesize the de-noised audio signal we rely on an iterative pre-image method. In order to gain better understanding about the pre-image step we performed experiments with different pre-image methods, first on synthetic data and then on audio data. The results of these experiments led to a reduction of artifacts in the original speech enhancement method, tested on speech corrupted by additive white Gaussian noise at several SNR levels. The evaluation with perceptually motivated quality measures confirms the improvement.