Signal time-frequency decomposition based sound data classification method

Abstract

The invention discloses a signal time-frequency decomposition based sound data classification method. The method comprises that a detected metal tank product is excited to vibrate; sound generated byvibration of the detected project is collected and sampled to obtain a digital sound signal; the sound signal is processed to obtain a frequency spectrum signal of the sound signal; the sound signal and the frequency spectrum signal thereof are decomposed to obtain a time-domain characteristic matrix W1 and a frequency-domain characteristic W2 that can represent the product quality respectively; {W1, W2} as characteristic data of the product quality are input to a product quality calculation model for operation, and an operation result of the module refers to a product quality parameter R; anabsolute value D of a difference between R and a quality standard value delta of the qualified product is calculated; and if D is lower than a threshold T, the detected product is qualified, and otherwise, the detected product is unqualified. The quality of the metal tank product is detected on the basis of a sound emission technology, and the detection correctness is improved effectively.

Description

technical field [0001] The invention relates to the field of quality detection of metal canned products based on acoustic emission technology, and more specifically relates to a sound data classification method based on signal time-frequency decomposition. Background technique [0002] Metal cans are generally used as containers to package products, which can play a role of sealing and aesthetics, and are widely used in the food industry. However, unqualified metal canned food is likely to cause food quality problems. Once consumed by consumers, it is very easy to cause Hazardous situations such as illness or poisoning. Although almost all metal canned products are subject to quality inspection before leaving the factory, due to some products only undergoing random inspection, or missed inspections or false inspections during the full inspection process, a small number of defective products are not detected and flow to the market. Therefore, it is necessary to study an accu...

AI Technical Summary

Problems solved by technology

[0011] The above three schemes only use the frequency domain information of the sound signal, especially the frequency domain peak value and its frequency as the basis for judging the quality of the inspected product, while the deformation and defect of the excited vibration part of the metal canned product, and The interference of random noise will affect the frequency spectrum of the sound signal generated by its vibration, which may lead to misjudgment of product quality

Images