Abstract:There is a critical need to develop and validate non-invasive animal-based indicators of affective states in livestock species, in order to integrate them into on-farm assessment protocols, potentially via the use of precision livestock farming (PLF) tools. One such promising approach is the use of vocal indicators. The acoustic structure of vocalizations and their functions were extensively studied in important livestock species, such as pigs, horses, poultry and goats, yet cattle remain understudied in this context to date. Cows were shown to produce two types vocalizations: low-frequency calls (LF), produced with the mouth closed, or partially closed, for close distance contacts and open mouth emitted high-frequency calls (HF), produced for long distance communication, with the latter considered to be largely associated with negative affective states. Moreover, cattle vocalizations were shown to contain information on individuality across a wide range of contexts, both negative and positive. Nowadays, dairy cows are facing a series of negative challenges and stressors in a typical production cycle, making vocalizations during negative affective states of special interest for research. One contribution of this study is providing the largest to date pre-processed (clean from noises) dataset of lactating adult multiparous dairy cows during negative affective states induced by visual isolation challenges. Here we present two computational frameworks - deep learning based and explainable machine learning based, to classify high and low-frequency cattle calls, and individual cow voice recognition. Our models in these two frameworks reached 87.2% and 89.4% accuracy for LF and HF classification, with 68.9% and 72.5% accuracy rates for the cow individual identification, respectively.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to address the following key issues: 1. **Developing Non-Invasive Indicators**: - The paper emphasizes the importance of developing and validating non-invasive indicators of animal emotional states to integrate these indicators into farm assessment protocols, especially when using Precision Livestock Farming (PLF) tools. 2. **Studying Vocal Characteristics of Cows in Negative Emotional States**: - Vocal communication behavior in cattle has been relatively understudied, particularly in negative emotional states. The paper attempts to fill this knowledge gap by collecting and analyzing vocal data from cows in negative emotional states, especially during visual isolation challenges. 3. **Classifying High-Frequency and Low-Frequency Calls**: - The paper proposes two computational frameworks (based on deep learning and interpretable machine learning) to classify high-frequency (HF) and low-frequency (LF) calls of cows to identify vocal characteristics under different emotional states. 4. **Individual Cow Vocal Recognition**: - In addition to classifying call types, the paper also explores methods for recognizing individual cows through their vocalizations, which helps further understand individual differences among cows. 5. **Improving Farm Animal Welfare**: - By monitoring cow vocalizations, scientists can better understand the emotional states of animals, leading to more informed decisions to improve farm animal welfare. ### Research Background - **Farm Animal Welfare**: Farm animal welfare is typically defined as the balance of positive and negative emotions. Recent research indicates that emotional states are crucial for the health and well-being of farm animals. - **Application of Bioacoustics**: Bioacoustics has been used to assess the health, emotional states, and stress responses of several important livestock species, such as pigs, goats, horses, and poultry. However, research on cows is relatively limited. - **Vocal Characteristics of Cows**: Cows can produce two types of vocalizations: low-frequency calls (LF) and high-frequency calls (HF). Low-frequency calls are usually associated with lower stress or positive emotions, while high-frequency calls are related to higher emotional arousal and negative emotional states. ### Research Methods - **Data Collection**: The research team collected vocal data from 20 adult lactating cows under controlled conditions in negative emotional states, totaling 1144 records, with 952 high-frequency calls and 192 low-frequency calls. - **Model Development**: Two computational frameworks were developed: one based on deep learning and the other on interpretable machine learning, for classifying high-frequency and low-frequency calls and recognizing individual cows. - **Performance Evaluation**: The models' performance was evaluated through cross-validation, showing that the two models achieved 87.2% and 89.4% accuracy in classifying high-frequency and low-frequency calls, respectively, and 68.9% and 72.5% accuracy in individual cow recognition. ### Conclusion - **Contribution of the Dataset**: This study provides the largest and cleanest dataset of cow vocalizations in negative emotional states to date. - **Effectiveness of the Models**: Experimental results indicate that the proposed deep learning and interpretable machine learning models perform well in classifying high-frequency and low-frequency calls and recognizing individual cows. - **Future Applications**: Monitoring cow vocalizations can provide valuable tools for assessing animal emotional states, promoting the development of precision livestock farming, and improving farm animal welfare. ### Future Work - **Expanding Research Scope**: More vocal data from cows need to be collected in various contexts, considering more potential influencing factors to gain a more comprehensive understanding of cow emotional states. - **Real-Time Monitoring**: Developing real-time monitoring systems for early disease diagnosis and health risk identification. Through these efforts, the research team hopes to provide new insights and technical support for improving the welfare and health of farm animals.

BovineTalk: Machine Learning for Vocalization Analysis of Dairy Cattle under Negative Affective States

BovineTalk: machine learning for vocalization analysis of dairy cattle under the negative affective state of isolation

Multi Modal Information Fusion of Acoustic and Linguistic Data for Decoding Dairy Cow Vocalizations in Animal Welfare Assessment

Affective State Recognition in Livestock—Artificial Intelligence Approaches

Classification of pig calls produced from birth to slaughter according to their emotional valence and context of production

Deep Learning-Based Cattle Vocal Classification Model and Real-Time Livestock Monitoring System with Noise Filtering

Dairy Cow rumination detection: A deep learning approach

Identification of oestrus cows based on vocalisation characteristics and machine learning technique using a dual-channel-equipped acoustic tag

Research on Automatic Recognition of Dairy Cow Daily Behaviors Based on Deep Learning

Classifying Ingestive Behavior of Dairy Cows via Automatic Sound Recognition

Monitoring the Respiratory Behavior of Multiple Cows Based on Computer Vision and Deep Learning.

Uncovering Patterns in Dairy Cow Behaviour: A Deep Learning Approach with Tri-Axial Accelerometer Data

Sound identification of abnormal pig vocalizations: Enhancing livestock welfare monitoring on smart farms

Video-based cattle identification and action recognition

Deep learning image recognition of cow behavior and an open data set acquired near an automatic milking robot

Dairy DigiD ~ A Deep Learning-Based, Non-Invasive Biometric Identification System for Dairy Cattle Using Detectron2

Decoding Poultry Vocalizations - Natural Language Processing and Transformer Models for Semantic and Emotional Analysis

The Use of Artificial Intelligence in Assessing Affective States in Livestock

A computer vision approach based on deep learning for the detection of dairy cows in free stall barn

C3D-ConvLSTM Based Cow Behaviour Classification Using Video Data for Precision Livestock Farming

New method for modeling digital twin behavior perception of cows: Cow daily behavior recognition based on multimodal data