Audio Features Extraction to Develop a Child Activity Recognition Model Using Support Vector Machine to Monitoring Security in a Smart City

Children activity recognition and classification is a subject of novel interest on which different works have been presented, where the data source used to classify the activities is determinant to define the other components of the classification model. This work uses environmental sound as data source to perform the recognition and classification of activities. A model for classifying children’s activities is presented based on the Support Vector Machine (SVM) algorithm, which is trained with features extracted from the audio samples of each of the activities defined for analysis: running, playing, crying and walking. As a result, a model able to predict, based on its contained information, the class to which a new sample of analyzed audio belongs is obtained.