论文部分内容阅读
为建立水下噪声音色特征的定量表达以用于目标识别,本文将主观评价实验获得的4个本质音色维度得分与声音的听觉中枢响应建立联系,得到音色的偏最小二乘回归模型,并基于回归系数对每个维度进行物理分析.为验证该方法的有效性,本文提取大量音色描述符作为自变量进行对比,结果表明听觉中枢模型预测能力有一定优势.同时发现,前3个本质音色维度可分别由高频能量比例、谱平坦程度和时域连续性描述,而第4维度则无法与任何声学特征建立联系.
In order to establish a quantitative expression of underwater noise color features for target recognition, this paper correlates the four intrinsic tone color dimension scores obtained by subjective evaluation with the auditory central response of sound to obtain the partial least-squares regression model of the sound, Regression coefficients for each dimension of the physical analysis.To verify the effectiveness of the method, this paper extracted a large number of tone descriptors as independent variables compared to the results show that the auditory central model has some predictive ability.At the same time found that the first three essential tone color dimension It can be described by high-frequency energy ratio, spectral flatness and temporal continuity, respectively, while the fourth dimension can not be associated with any acoustic feature.