点击次数：

发表时间：2022-10-10

发表刊物：大连理工大学学报

期号：5

页面范围：729-732

ISSN号：1000-8608

摘要：A nonstationary sequence predictor for missing audio signal recovery is presented and designed. The predictor consists of a few BP networks and a uniform DFT filter bank. Through the improved method of sample processing and designing every BP network respectively, the predictor reaches a high precision comparatively. The multirate parallel processing reduces the data quantity processed by one BP network and increases the speed of missing audio signal recovery. Stimulation results show that this method is effective and efficient.

备注：新增回溯数据