地震前虎皮鹦鹉(Melopsittacus undelafus L.)叫声的模式识别

近年来,我国多数地震动物观察站(SAOS)监测虎皮鹦鹉(BG=budegerigar)的叫声频度与地震的关系,并取得了一些观测结果。但是,所用的声记录仪未能考虑BG叫声的声学特性,而不能分辨BG不同的叫声和进一步提取震兆性叫声模式(EPCP),并易受到环境噪声的干扰而产生误计数。 据此,分析了BG不同声行为叫声的声学特性。BG叫声大致上可分为单次叫声(SC)、变音调叫声(VTC)、单音节和多音节抗议叫声(MoSPC和MuSPC)。由此,进一步研制了它们的模式识别系统。该系统是以单片机为核心的微处理机,对BG叫声进行预处理、编码、分析、显示和打印输出。 观察结果表明,在高声场旅游飞机噪声(TAN)干扰下,该系统的BG叫声识别率(CRR)为66.7％,而对TAN没有误计数。在含10次不同声强的TAN干扰(432秒)的30分钟内,该系统记录了588次叫声(SC、VTC、MoSPC和SPC分别为377、22、20和169次),仅丢失65次叫声,其CRR平均为90％。但是,SAOS目前使用的声记录仪,这10次TAN干扰所产生的误计数可能约为BGU叫声的2.6—13倍。 由此可见,这种模式识别系统不仅可满足于BG不同叫声的识别,且具有良好的抗噪声功能。这就使之可为地震前BG声行为观测提供更有用的信息。不仅可给出BG叫声频度的昼夜节律变化与地震的关系,并可为揭示BG的EPCP提供一定的依据。

PATTERN RECOGNITION OF CALLS OF BUDEGERIGAR (Melopsittacus undelafus L.) PRIOR TO EARTHQUAKES

On the basis of studies of the acoustic behavior of the budegerigar (BG), the pattern recognition of different calls of the BG is further given in this paper.The results show that under the disturbances in the tourist aeroplane noise (TAN) with different sound intensities, it is not only able to recognize the single call, vari-toned call, mono-syllabic and multi-syllabic protest call of the BG, and the recognition rate is 90% on an average, but did not produce erroneous counts in the TANs. In the sound recorders with different sampling rates employed now by seismological animal observation stations, the number of erroneous counts produced by the TANs may be about 2.6-13 times of BG calls.It may be seen that the pattern recognition of BG calls is better possessed of the specificity and anti-noise function. These enable it to provide more useful information for the acoustic behavioral observation in the BG before earthquakes, especially to extract the earthquake premonitory call pattern in the BG.