隐马尔可夫模型在舰船噪声目标识别中的应用  

Ship-radiated noise recognition based on CHMM

作　　者：戴卫国[1] 张宝华 程玉胜[1] 

机构地区：[1]海军潜艇学院,青岛266071 [2]解放军某部,北京100841

出　　处：《声学技术》2008年第2期187-191,共5页Technical Acoustics

摘　　要：将Mel频率倒谱作为特征参数,实现了基于连续型隐马尔可夫模型的舰船辐射噪声目标识别。对五艘不同类别舰船的辐射噪声建立模型,并进行了识别实验。模型的状态数和高斯混合数分别确定为1和3,五艘舰船的平均识别率达到91.6%。利用两艘不同类型舰船的噪声分析了舰船的工况对识别结果的影响。Ship-radiated noise recognition based on Continuous Hidden Markov Models（CHMM） is achieved. MelFrequency Cepstral Coefficients （MFCC） are used as features. Ship-radiated noises of five different classes are modeled and target recognition experiments are conducted, the number of states and Gaussian mixture is set to be 1 and 3 respectively. The average recognition rate is up to 91.6% for 5 different ships. Noises of two different ships are used to analyze the impact of different operating conditions on the recognition rate.

关 键 词：舰船辐射噪声 训练 识别 隐马尔可夫模型 MFCC 

分 类 号：O429[理学—声学]