青鳉鱼的行为特征提取研究

Behavior Feature Extraction Based on Medaka

采用生物行为传感器获取青鳉鱼的行为数据时,青鳉鱼个体差异会导致采集到的原始电信号在时空特性下完全不同。重要的行为特征往往被隐藏在原始信号中,传统的信号处理方法无法实时而有效地提取到这些特征。针对这个问题,观察并记录了暴露实验前后青鳉鱼的行为变化,提出了一种可以高效表征行为特征的直方图统计算法。实验结果表明,该方法能够准确对应人眼观测到的暴露实验前后鱼的行为变化趋势,同时也为后续异常行为识别提供一定的支持和参考。

When biological behavioral sensors were used to obtain the behavior data of the medaka,the individual differences of the medaka would cause the original electrical signals collected to be completely different under the temporal and spatial characteristics.Important behavioral features are often hidden in the original signal,traditional signal processing methods cannot effectively extract these features in real time.In this view,we observe and record the changes in the behavior of the medaka before and after the exposure experiment,and propose a histogram statistics algorithm,in order to effectively characterize the behavioral characteristics of the medaka.The experimental results show that this method can accurately reflect the changes before and after the exposure experiment observed by human eyes,and it also provides some support and reference for subsequent abnormal behavior identification.