METHOD FOR THE RECOGNITION AND AUTOMATIC CLASSIFICATION OF DOMESTIC DOG VOCALIZATIONS
The present disclosure is related to a method for detecting and classifying dog barks. The method captures the audio signal, detects periods where there are barks present, then acoustically characterizes them in audio segments looking for the best features by comparing a series of indicators with th...
Saved in:
Main Authors | , , , , |
---|---|
Format | Patent |
Language | English Spanish |
Published |
12.06.2018
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The present disclosure is related to a method for detecting and classifying dog barks. The method captures the audio signal, detects periods where there are barks present, then acoustically characterizes them in audio segments looking for the best features by comparing a series of indicators with those that are previously determined in a database. Automatic recognition models are trained to generate valence estimates and activation of each bark, a class is assigned according to the level of activation and valence reflected by the dog in his vocalizations. The system gives as an output the classification of the barks, which are compared with lists of diffuse attributes as matrices with sub processes.
La invención consiste en método para la detección y clasificación de ladridos. El método captura la señal de audio, detecta los periodos donde existen ladridos luego los caracteriza acústicamente en segmentos de audio buscando las mejores características comparando una serie de indicadores, con aquellos que se encuentran previamente determinados en una base de datos se entrenan modelos de reconocimiento automático para generar estimaciones de valencia y activación de cada ladrido, se le asigna una clase de acuerdo al nivel de activación y de valencia reflejado por el perro en su s vocalizaciones el sistema entrega como salida la clasificación de los ladridos, los cuales se comparan con listas de atributos difusos a manera de matrices con subprocesos. |
---|---|
Bibliography: | Application Number: MX20160016763 |