倒谱和功率谱分析在禽蛋破损检测中的应用

Application of Power Spectrum and Cepstrum to Detection of Cracked-shell Eggs

为了在蛋品加工中能够快速地检测并剔除破损蛋,将声卡采集到的蛋在敲击时的声脉冲信号送入计算机,利用倒谱和功率谱进行分析,分别提取出6个特征参数,并进行逐步判别分析,选择出最优判别因子,建立判别模型。实践表明:其对破损蛋的识别准确率达到95%,对禽蛋破损检测提供了参考方法。

In order to quickly find out and pick out cracked-shell eggs in the processing of eggs, the acoustics signals created by striking the eggshells are acquired by a sound card and transformed into digital signals. Through analyzing those signals, six characteristic parameters were extracted respectively from power spectrum and cepstrum. Step discriminance was used to optimize the combination of these characteristic parameters. Establish the optimum distinguishing functions of normal eggshells and cracked eggshells respectively. The accurate recognition ration for cracked-shell eggs is up to 95.5 percent.