Real-Time Analysis of Beats in Music for Entertainment Robots

The dancing actions for entertainment robots are usually designed in advance and saved in a database according to the beats and rhythm of the given music. This research is devoted to developing a real-time algorithm that can detect the primary information of the music needed for the actions of entertainment robots. The computation of the proposed algorithm is very efficient and can satisfy the requirement of real-time processing by a digital signal controller. The digitized music signal is first normalized to make the algorithm robust for miscellaneous music volume. Next, the primary features of the beat for the music are extracted. Based on the extracted features, the algorithm will then identify the occurrence of beats in the music through the use of an optimized classifier. From the accumulated information of the timing for the beats, the algorithm can provide the predicted timing information of the next beat through regression analysis. The type and tempo of the given music can also be derived thereafter. The entertainment robot can thus move or dance according to the information derived by the algorithm. A graphical user interface (GUI) program in LabVIEW is also utilized to observe and verify the analysis results. In this study, the ratio for correct beat detection is greater than 90%. The prediction ratio for the correct timing of beats is over 80%, and it is 100% correct for both music type and music tempo.