DEEP LEARNING-BASED INTEGRATED SOUND INFORMATION COGNITION SYSTEM

摘要

The present invention relates to a deep learning-based integrated sound information cognition system. More specifically, the deep learning-based integrated sound information cognition system comprises: a sound input module to which a sound is inputted; a sound event detection module for dividing a sound signal inputted from the sound input module into frames of a predetermined first time unit, and detecting a sound event on the basis of analyzed frame information; and an environment sound detection module for dividing a sound signal inputted from the sound input module into frames of a predetermined second time unit (which is a longer time unit than the first time unit), and detecting an environment sound on the basis of analyzed frame information. According to the deep learning-based integrated sound information cognition system of the present invention, a sound signal inputted from the surroundings is analyzed by being divided into short time units to detect a sound event, and is analyzed by being divided into longer time units to cognize an environment sound. A surrounding environment is cognized on the basis of two integrated pieces of information, so a surrounding environment can be more accurately cognized.;COPYRIGHT KIPO 2017