Guitar chord recognition based on finger patterns with deep learning

Many guitar players use video contents such as Youtube to practice. If the content contains noise or background sounds, then the player must watch the videos repeatedly, which is very troublesome. In order to solve this problem, we attempt to build a system that can recognize the finger patterns of guitar players in video and can automatically generate a corresponding musical score. The present paper introduces a method to recognize finger patterns with deep learning. Experimental results reveal that a three-chord classifier can achieve a recognition rate of approximately 90% and a five-chord classifier can achieve a recognition rate of approximately 70%.