Research And Application Of Real-Time Music Beat Tracking System

With the development of computer technology, computer applications continue to penetrate into all levels of production and life of the people. As an indispensable art, which can be taken as an expression of people's thoughts and feelings, reflection of people's real life, music is paid more and more attention by computer scholars. People aware that applying computer technology into the field of music can greatly enrich the expressiveness of music and it have a very broad application prospects. Augmented reality technology is a hot study of computer technology in recent years, whose core idea is to combine computer-generated virtual objects with real scenes to achieve the purpose of integration of the actual situation, which is one problem of the study of computer vision. How to apply the music information into the augmented reality environment of computer technology effectively is a novel problem, for example, we can drive the virtual objects in the virtual reality environment by identifying the beat of music. This leads to the concept of music beat tracking.Music beat tracking is a process when people listen to music they can clap and stomp along with the beat. Currently, the problem of real-time music beat tracking faces many challenges. First, the real value of the rhythm is not defined, which can only be determined in accordance with people's feelings and be tried to close people's perception of the music. This is extremely easy for people, while it is a challenging task for computer. Second, the obtain of people's annotation may have some difficulties. Third, the evaluation method also has many difficulties. First of all, although many evaluation methods have been proposed by researchers, it did not reach a consensus, so there is no uniform standard. Because the input and output may have different expression forms among different algorithms and also their applications are different. Once more, many data sets used by researchers are very small and proprietary, which can make the comparison of different algorithms more difficult. Finally, different types and different complexity of music often led to difficulties of evaluation.Most of the beat tracking algorithms proposed before have some limitations in some degree, so we presents a wavelet-based real-time music beat tracking algorithm, which can detect the beat value and the specific beat locations of the music in real time and is compared to the result provided by MIREX 2006 for using their data sets and evaluation method. The experimental results show that our algorithm ensures the accuracy and also improves the performance of time, for other algorithms are non-real-time and ours is real-time. In addition, we applied our beat tracking algorithm to augmented reality.