A broiler overlap sound recognition method based on confidence interval

By identifying overlapping sounds in broilers using a confidence interval-based method, the problem of low accuracy in recognizing overlapping sounds in broilers in existing technologies is solved, enabling efficient broiler health monitoring and improving recognition accuracy and sensitivity.

CN116564317BActive Publication Date: 2026-06-12HEILONGJIANG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HEILONGJIANG UNIV
Filing Date
2023-05-19
Publication Date
2026-06-12

AI Technical Summary

Technical Problem

Existing methods have low accuracy in recognizing overlapping sounds in broilers, which affects the assessment of broiler health status.

Method used

A confidence interval-based approach is adopted, which uses a machine learning model classifier to perform frame processing and endpoint detection on the broiler sound signal, extracts frame signal feature data, determines the sound segment category by confidence rate, and sets different confidence intervals to identify individual sound categories and overlapping sounds, including single, double, triple overlapping sounds and their combinations.

Benefits of technology

It achieved accurate recognition of overlapping sounds in broilers, improving recognition efficiency and the sensitivity of health monitoring. The recognition accuracy reached 97.22%, with an error rate of 2.27%, which improved the accuracy of cough rate calculation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116564317B_ABST
    Figure CN116564317B_ABST
Patent Text Reader

Abstract

The application discloses a broiler overlapping sound recognition method based on a confidence interval, and relates to a broiler overlapping sound recognition method. In order to solve the problem that the existing method has low recognition accuracy in recognizing the overlapping sound of broilers, the broiler sound signals collected are subjected to frame processing and endpoint detection, and sound feature data are extracted for each frame signal; a machine learning model classifier is used to predict each sound feature data in each sound feature data set, calculate the first confidence rate of each sound feature data set, and judge whether the first confidence rate of each sound feature data set is located in the [beta1, 1] interval, wherein beta1 is determined based on the lower limit value of the union of the first confidence intervals corresponding to different single sound categories; if yes, it is determined that the corresponding sound segment is a single sound category, and the label of the maximum data number of the calculated first confidence rate is determined as the sound category corresponding to the sound segment; otherwise, it is determined that the corresponding sound segment is overlapping sound.
Need to check novelty before this filing date? Find Prior Art