ANIHWA Toward a tool for farmers to evaluate welfare states of pigs: measuring vocal indicators of emotions

Hovedmålet med SoundWel var å utvikle et verktøy som kan brukes for å måle emosjonelle tilstander hos gris basert på deres vokalisering. For oppgave 1.1 skulle alle prosjektpartnere lage en database med vokaliseringer fra gris i forskjellige sammenhenger man mener oppleves negativt eller positivt. Databasen er nå på plass og har gjort det mulig å identifisere områder som manglet slik at vi kunne foreta tilleggsregistreringer. For oppgave 1.2 der Norge, Frankrike og Tsjekkia hadde ansvaret, har vi foretatt tilleggsregistreringer for vokaliseringstyper som mangler i databasen ved å fokusere på situasjoner som grisen opplever som positivt. For oppgave 1.3. der NO og Sveits hadde hovedansvaret har vi gjennomført meta-analyse av vokaliseringer i databasen for å identifisere egenskaper som gjør det mulig å skille forskjellige emosjonelle tilstander hos grisene eksponert for forskjellige situasjoner. For oppgave 2.1. har alle deltagere bidratt til å definere et system som kan brukes for å detektere emosjonelle tilstander hos gris på grunnlag av vokaliseringer. I tillegg skulle alle prosjektdeltagere i oppgave 2.4. få opplæring i bruk av et nytt system som kan brukes for å detektere emosjonelle tilstander hos gris på grunnlag av vokalisering. Grunnet uforutsette forhold var det ikke mulig å lage det nye systemet eller få trening i bruken. En tidligere versjon av software for analyse av vokalisering som heter STREMODO ble brukt av prosjektdeltagerne, men krever mer omfattende opplæring.

Outcomes have involved sharing data and expertise that enabled the production of a unique database that now forms the basis for defining a system allowing reliable automated detection pigs emotions through future projects. Several young researchers were trained through the project, enabling them for future research careers within this field. The project facilitated interdisciplinary collaboration between specialists in sound analysis, computer engineering, programming, and ethology. In the long-term the results and competence generated through this research will enable the production of system that will allow producers to monitor the welfare of their pigs at the farm level, thus addressing the societal demand for agricultural products meeting high ethical standards. In the short term the project group continue to work together both with the aim of developing a prototype system for detecting emotions in pigs and with the aim of attracting funding for further work in this area.

The importance of the impact of mental health on animal welfare has become evident, and animal welfare is now assessed through both physical and mental health. However, tools to scientifically measure the affective lives of animals are still lacking. One promising tool to assess animal emotions is through vocalisations. Our project proposes to develop a robust system for identifying emotions in fattening pigs using vocalisations, which could be used as welfare indicators on-farm. This will be done by gathering European experts on vocal indicators of mental states in pigs (FR, DE, CH, NO, CZ). The project includes three main tasks. The first task aims at identifying vocal indicators of emotions in pigs. This will be done with a meta-analysis on existing and new data recorded within the project. The groups already has a large number of audio recordings from situations associated with different emotions, ranging from stressful procedures (e.g. castration and handling) to positive situations (e.g. social reunion and post-nursing interactions). Because the recordings vary in terms of breed, sex, age and recording techniques, we will also carry out more recordings in standardised conditions, in order to obtain a baseline for the recognition system, and in missing situations (positive states and slaughter). The second task aims at developing a recognition tool able to identify emotions using vocalisations. It will consist in testing different classification methods/models, in order to choose the more appropriate to develop a system that could reliably identify pig emotions. Then, the recognition software will be built. The third and last task will consist in testing and validating the system. To this aim, the teams will test it in various situations at their own research stations and on-farm (from birth to slaughter). The final outcome of this project will be an accurate non-invasive system for identifying pig instantaneous emotional states.