An Efficient Hidden Markov Model with Periodic Recurrent Neural Network Observer for Music Beat Tracking

In music information retrieval (MIR), beat tracking is one of the most fundamental tasks. To obtain this critical component from rhythmic signals, a previous system hidden Markov model (HMM) with recurrent neural network (RNN) observer was developed. Although frequency quite stable, existing HMM based methods do not take feature into account. Accordingly, states in these HMM-based are redundant, which disadvantage for time efficiency. paper, we proposed an efficient using by exploiting contents network’s observation Fourier transform, extremely reduces computational complexity. Observers that works used, such as bi-directional (Bi-RNN) and temporal convolutional (TCN), cannot perceive beat. more reliable frequencies music, periodic (PRNN) on attention mechanism well, used HMM. Experimental results open source datasets, GTZAN, Hainsworth, SMC, Ballroom, show our PRNN competitive to state-of-the-art has lower cost.