A New Type of Speech Down-sampling Coding with High Frequency Estimation using Neural Networks - Speech Down-sampling Coding with Aliasing Noise
Hideyuki Urita, Kiyoshi Mizui · International Symposium on Information Theory and its Applications · 1994
In this paper, we propose a new type of Speech Down-Sampling Coding with High Frequency Estimation using Neural Networks. The proposed type doesn't have an anti-aliasing filter for down-sampling. The Neural Networks in the receiver estimates the high frequency range from the low one from transmitter with the aliasing noise. This system can be used with any other coding methods because it uses only down-sampling for data reduction. Therefore, it has possibility of lower transmission rate. From results of computer simulation, it is confirmed that the proposed type with aliasing noise is better than the conventional one without it.