Music AnalysisAs demonstrated in our publications or our yearly participation in the MIREX evaluation campaign, we possess state-of-the-art methods for several general music analysis tasks, some of which are listed below.
What do you have?
Onset and Beat Detection
We can detect the starting points of all notes played on any instrument or sung, as well as drum hits, both in solo recordings and polyphonic music pieces (be it popular or classical), with state-of-the-art accuracy (see MIREX 2013 results, algorithm SB1). We can also detect and track the beat in music pieces with state-of-the-art accuracy (see MIREX 2013 results, algorithm SB6).
We can determine the tempo of a recording in beats per minute with state-of-the-art accuracy (see MIREX 2013 results, algorithm SB7).
We have a reasonably accurate algorithm for detecting the onset times and pitches of piano notes (see Böck et al.). This method is also backing some of our music identification and alignment technologies.
We can analyze a music piece to give it a number of textual labels from a previously learned repertory, such as guitar, slow, romantic, british, with state-of-the-art accuracy (see MIREX 2012 results, algorithm SSKSS1).
Description of Textural Quality
We have determined dimensions used by humans to describe sound textures, such as high/low and ordered/chaotic, then developed a method to obtain the strength of these dimensions from analyzing an audio file and a method for visualizing the description (try a browser demo and see Grill et al. for details).