Acoustic feature extraction and optimization of crack detection for eggshell

• We constructed a system for eggshell crack detection in acoustic method.
• Features were extracted and optimized before recognition system.
• Features have high F-ratios and low correlations with any other feature are reserved.
• Classification accuracy of different impact positions and crack positions were measured.
• Higher efficiency and processing speed can be received in recognition stage.

Research on egg crack detection has been conducted based on acoustic method, especially those on the extraction of different features and those that compare sound and cracked eggs. Thus, an excitation device that is driven by solenoid was developed in the current study to generate acoustic signals by impacting an egg. Time and frequency domain features that were used and customized by previous researchers and customized were extracted. F-ratio was used to evaluate the effect of every feature on the discrimination of sound and cracked eggs, and correlations between features were investigated. Features with remarkably low F-ratio values and those with relatively low values in pairs of highly correlated features were disregarded. Classification accuracy reach 99.2% using neural network with features reduction in this method and features reduction will assist in simplifying recognition algorithms and in reducing computations in on-line systems.