四川大学学报（自然科学版）Issue(2)：292-298,7.DOI:103969/j.issn.0490-6756.2014.02.014
一种听觉显著图提取模型
A saliency map extraciton model for auditory attention
摘要
Abstract
A new bottom-up saliency map for modeling auditory attention is proposed based on the char-acteristics of human auditory.Saliency maps can be effectively obtained by our model which is based on the two facts that auditory signal is frequency-divisionally processed by human auditory and spectrogram can be treated as a single dimensional temporal input at all times.First,the spectrogram is formed by applying the SFFT to the auditory signal.Then,the auditory spectrogram is divided into several fre-quent bands,and different sets of one-dimensional Gaussian filters are used to get different levels.The saliency maps at different scales at each band are achieved by center-surround mechanism.And by com-bining maps acorss scales,the saliency map for each specific frequency band is determined.The final sali-ency map for auditory attention is obtained by the linear combination of saliency map at each frequency band.The simulation results of auditory saliency map from the new model proposed are proven to per-form more accurately than the previous.关键词
听觉显著图/显著图提取模型/听觉注意Key words
Auditory saliency map/Saliency map for modeling attention/Auditory selective attention分类
信息技术与安全科学引用本文复制引用
张欣,夏秀渝,王雪君..一种听觉显著图提取模型[J].四川大学学报（自然科学版）,2014,(2):292-298,7.基金项目
国家自然科学基金(61071159) （61071159）
