In order to estimate numbers of single species freshwater fish,the passive acoustic signals of different quantities crucian carp and bream fish were used as samples.The acoustic signals of the freshwater fish were collected by the HTI-96-MIN standard hydrophone.The extracting schemes of sample characteristics were compared by using 4 ～ 6 layer wavelet packet decomposition algorithms.The short-time average zero-crossing rate and the frequency band energy calculated by 6 layer wavelet packet decomposition algorithm were determined as the signal characteristics of crucian carp.And the characteristics of bream fish were the frequency band energy calculated by 6 layer wavelet packet decomposition algorithm only.According to the research results of different sample sets dividing methods,the final sample sets dividing methods of crucian carp and bream fish were Rank-SPXY (m =10) and Rank-SPXY (m =5).Furthermore,the key characteristics of the samples were chosen by competitive adaptive reweighted sampling.Finally,some numbers estimate models of the freshwater fish were established with the multiple linear regression and partial least squares regression.The results indicated that the performance of the multiple linear regression model was better.The multiple correlation coefficients of the numbers estimate models of crucian carp and bream fish were 0.835 and 0.893,and the relative percent deviations were 1.79 and 2.Ol,respectively.The numbers estimate methods proposed had good performance.It provided a new means for monitoring the freshwater fish quantity.%针对单品种淡水活鱼数量估计问题,以鲫鱼和鳊鱼为研究对象,采用不同数量淡水鱼的被动水声信号作为样本,通过4～6层小波包分解算法,对比分析了样本的若干特征提取方案,明确了鲫鱼水声信号特征的提取方案为短时平均过零率+6层小波包分解频段能量,而鳊鱼只需提取6层小波包分解频段能量作为其信号特征.根据不同样本集划分方法比较结果,确定了鲫鱼样本集划分方法为Rank-SPXY(m=10),鳊鱼为Rank-SPXY(m=5).利用竞争自适应重加权采样,筛选出样本的关键特征,并使用多元线性回归和偏最小二乘回归建立了若干淡水活鱼的数量估计模型,结果表明:多元线性回归模型的性能较好,鲫鱼和鳊鱼数量估计模型的复相关系数分别为0.835和0.893,相对分析误差分别为1.79和2.01.
展开▼
