泣き声による児童虐待検知装置の提案

Abstract

In recent years, the number of cases on child abuse was increased. Eliminating child abuse and preventing children from child abuse are needed. In order to reduce child abuse, ongoing monitoring on abused children is awaited, which is without case workers. Therefore we propose the monitoring system for abused children, as a wearable computer which detects child's crying. We propose both software algorithm and embedded hardware system. First, the method of cry detection by sound is proposed. The sound of cry consists of 2 major mode, cry with words and cry without words. The latter presents a melody, a same pitch keeps with some periods. Its pitch varies in order. Therefore, variation of pitch shows the period of melody cry. Experimental results shows its effectiveness. Next, we approach the hardware implementation. ARM based processor with DSP will execute extraction method of prosodic features. We tested the microphone control using microprocessor, and con rmed the captured data can be treated as a sound data.