Evaluation method for quality losses generated by miscellaneous audio signal processings using MIDI encoder tool “Auto-F”

We have been developing a MIDI encoder tool, which can convert given general acoustic signals to MIDI codes. With these converted codes, we can reproduce an acoustic signal similar to the original using a standard MIDI sound module. In this paper, we propose applying our proposed MIDI encoding technique to comparative evaluation of similar acoustic signals such as some processed signal and its original one. In order to apply this tool to quality evaluation of acoustic signals, we have improved temporal resolution in a frequency analysis process. More specifically, we have improved a variable analysis frame shifting algorithm and temporal grouping precision of analyzed components. Moreover, proposing creating a set of differential MIDI data generated by an exclusive-or operation between two sets of encoded MIDI data, we can recognize differences between given two acoustic signals either visually or audibly. In this paper, we describe an abstract of our improved MIDI encoding algorithm. And using this improved tool, we present trials of both visualization and auditorialization of quality losses, especially made by audio watermark embedding processes.