AUDIO TEXTURE SYNTHESIS METHOD USING EM OPTIMIZATION

• Main Authors: ROE, CHANG HWAN, YOO, MIN JOON, LEE, IN KWON
• Format: Patent
• Language: English
• Published: 05.08.2008
• Subjects: ACOUSTICS; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

An audio texture synthesis method using EM optimization is provided to synthesize an example audio directly by using an optimization method to reduce repetition and generate audio having a desired length, thereby increasing user's immersion in an interactive environment such as a game. An audio texture synthesis method using EM optimization comprises the following steps of: allowing an equalization unit to equalize the amplitude value data of an example audio texture through RMS(Root Mean Square) if the example audio texture is inputted to an example audio input unit(20,30); allowing the result generator of an EM optimization unit to divide the received amplitude value data by predetermined unit, synthesize divided data to be overlapped with each other as long as a predetermined length, and generate an optimized result audio texture of an arbitrary length(41); allowing the comparison unit of the EM optimization unit to compare the generated result audio texture with the example audio texture and search a part similar to the result audio texture in the example audio texture(43); and generating a final result audio texture(50,60).