Discrimination strategy using machine learning technique for oestrus detection in dairy cows by a dual-channel-based acoustic tag.

•A neck tag equipped with low-cost dual-channel microphones and a multi-threshold-based sound extraction method were developed to effectively collect vocalisations of individual cows.•A series of statistical methods consisting of significance analysis, correlation analysis, and individual difference analysis were adopted to evaluate the availability of acoustic and statistical features of dairy cow vocalisations for heat detection.•A dual-LSTM joint discriminant strategy based on optimal parameters was proposed to achieve the identification of oestrus state and the estimation of oestrus onset.