The Research On Piano Music Dereverberation Technology

The reverberation of recorded piano signal will be more obvious when the distance of the audio source is far away from the microphone in relatively enclosed space, such as conference room、concert hall etc. The existence of reverberation will make the sound confused, difficult to distinguish, and declining the quality of sound, furthermore it reduces the melody extraction and retrieval quality of piano music to a great extent, and affects the accuracy of piano recognition seriously. Therefore, dereverberation is necessary before melody extraction and retrieval of piano music. So, dereverberation of piano music is a very important research subject. It is very important significance to melody extraction, retrieval of piano music and people’s entertainment life.The dereverberation algorithm based on a two stage dereverberation algorithm of speech is improved. Firstly, piano music reverberation is synthesized by researching and analyzing several typical digital Reverb models. Then Onset detection algorithm is introduced, based on the difference of pronunciation principles of piano music and speech. That is, piano music note segmentation should be completed before removing late reflections of piano music. Then spectral subtraction is employed to remove late reflections of piano music segmentally.The main works are as follows:1、The first stage dereverberation algorithm of a two stage dereverberation algorithm of speech is applied to piano music signal, and the optimal adaptive inverse filter that removing early reflections of piano music is obtained, then we make some experimental simulation, The results showed that the first stage dereverberation algorithm reduced the strength of piano music reverberation.2、The second dereverberation algorithm based on a two stage dereverberation algorithms of speech is improved. First of all, two onset detection algorithms is analysed: One is based on signal’s spectral difference and the other is based on RTFI. Comparing detection results of the two algorithms, the results showed that the algorithm which is based on RTFI could receive better effect. Due to the late reflections which one note affect on itself is considered only, therefore, after the completion of onset detection, the second stage dereverberation algorithm (spectral subtraction) is employed to remove late reflections of every two note,then make some experimental simulation, The results showed that the improved second stage dereverberation algorithms enhanced the clarity of piano music’s spectrograms.3、Several commonly used dereverberation algorithms of speech is applied to piano music, and the simulation results between the proposed algorithm and several algorithms in common use is compared, meanwhile subjective and objective evaluation criterias is employed to evaluate and analyse the experimental results. The results showed that the proposed algorithm increased the dereverberation degrees, enhanced the quality and clarity of piano music. Finally, the improved algorithm is used to dereverberation for different rhythm piano music. The paper drew the conclusion that the improved algorithm is more suitable for slow rhythm piano music.