针对常规人工记录锤击数的低效情况,在此将打桩现场采集的声音信号运用快速傅立叶变换到频率域,对频谱进行多个统计量的分析,并同时考虑了锤击声音可能受到附近噪音的干扰,从中选取平均绝对值偏差这个统计量作为识别锤击次数的特征,由此探测得到一组可以较为精确识别锤击数目的特征参数,实现了锤击数准确识别的自动化。
On the basis of the analysis of fast Fourier transform and some basic statistical variables, the piling sound is analyzed in frequency domain; then the character, mean absolute deviation, and corresponding parameters on the piling sound recognition are acquired; thus an algorithm can be developed to count the number of piling sound by rule and line.