Catálogo de publicaciones - tesis
Título de Acceso Abierto
Reconocimiento de actividades a partir de señales inerciales y acústicas
Sebastián Rodrigo Vanrell Hugo Leonardo Rufiner Omar Chiotti Pablo Mandolesi Humberto Torres Mariano Rubiolo Diego Humberto Milone
acceptedVersion.
Resumen/Descripción – provisto por el repositorio digital
Activity recognition aims at identifying the activities that a person or animal performs over time. Several challenges are faced in the discrimination of activities, such as processing long-term recordings and dealing with information given at different timescales. This thesis deals with automatic activity recognition from inertial and acoustic signals, which are captured with sensors on the body of a person or an animal. Inertial signals indirectly capture body movements and are used in both humans and dairy cows. Regarding human activities, a new representation of signals was studied, based on cepstral analysis, which captures the body dynamic that is associated with the activities of interest. Cepstral features perform very well in the recognition of daily activities and outperform state-of-the-art features. In a similar scheme, preliminary experiments showed that processing inertial signals helps to detect the onset of estrus in dairy cows. On the other hand, acoustic signals are used to monitor foraging behavior in dairy cows. An algorithm is proposed for long-term analysis of this behavior to recognize grazing and rumination blocks. In a first stage, a complete recording is analyzed, based on the autocorrelation of the sound envelope, to detect regular masticatory events and to define the time boundaries of activity blocks. In a second stage, the energy of acoustic signals within a block is analyzed to detect interruptions and characterize their regularity. Rumination blocks present regular interruptions, whereas grazing blocks do not. An extensive evaluation of the proposed algorithm is made and the results obtained are very good.Palabras clave – provistas por el repositorio digital
Activity recognition; Signal processing; Machine learning; Cepstral analysis; Precision livestock farming; Acoustic monitoring; Reconocimiento de actividades; Procesamiento de señales; Aprendizaje maquinal; Análisis cepstral; Ganadería de precisión; Monitoreo acústico
Disponibilidad
| Institución detectada | Año de publicación | Navegá | Descargá | Solicitá |
|---|---|---|---|---|
| No requiere | 2018 | Biblioteca Virtual de la Universidad Nacional del Litoral (SNRD) |
|
Información
Tipo de recurso:
tesis
Idiomas de la publicación
- español castellano
País de edición
Argentina
Fecha de publicación
2018-08-14
