Automated Acoustic Monitoring of Oestrus in Dairy Cows and Sows: Behavioural Evidence, Signal Processing, and Detection Methods

Oestrus detection determines insemination timing and reproductive efficiency in dairy cow and sow herds, yet manual observation and single-sensor systems remain vulnerable to weak behavioural expression and commercial barn conditions. This structured narrative review organises evidence across biological plausibility, acoustic acquisition, barn interference, feature representation, recognition models, multimodal fusion, decision support, and on-farm validation. Searches of PubMed and Crossref through 3 October 2026 produced 373 records and 339 unique records after deduplication. Of the core database records, 16 met the direct relevance criteria, and 78 supplementary biological, engineering, and methodological sources were incorporated through backward citation searching and targeted contextual retrieval, yielding 94 cited sources. Dairy cow evidence most consistently supports increased vocalisation rate around the oestrus climax and individual acoustic monitoring under defined conditions. Sow evidence depends more strongly on boar stimulation, the standing reflex, and multimodal labels, positioning acoustics mainly as an auxiliary behavioural cue. Reported direct recognition accuracies range from 91.25% to 98.92%, but heterogeneous labels and predominantly internal validation prevent statistical pooling or claims of commercial generalisability. Acoustic monitoring is best positioned as an early-warning and verification layer rather than a standalone predictor of ovulation or optimal insemination timing. Priority needs are standardised reference labels, leakage-resistant animal- and farm-level validation, device robustness, multimodal uncertainty handling, open benchmarks, and evaluation of reproductive and economic outcomes.