The blues genre of music was analysed by software that was created in this thesis. The results were consistent with the intuition that blues music does vary in key and tempo. Although not that many blues MIDI files were tested, the number was still high enough for the properties to show through.

Music generation was also studied in this thesis. An algorithm for generating new music based on guitar chords was presented and implemented using the Haskore Computer Music System. The main idea was to generate a chord progression, with notes from arpeggios forming the melody of the composition. Results obtained indicate that the algorithm provides some interesting compositions, that do not sound that unpleasant to listen to. Several stages helped to improve the compositions by adding a little more each time.

Future work

One of the limitations in this thesis was the assumption that a single time signature applied to the whole MIDI file. This assumption simplified the dividing of the MIDI file into bars. If, however, the MIDI file actually contained multiple time signatures, the number of bars would not be calculated correctly. Further research based on this thesis could modify the source code written to take into account multiple time signatures that a MIDI file may contain. Not only would this make the calculation of bars as accurate as possible, it would also increase the range of possible MIDI files that could be analysed by the software.

Guitar chords were used as the basis for generating new music. It would have been impractical to program all possible guitar chords because they number in their thousands. Not only that, but a particular guitar chord can have many different fingerings, or possible positions that the chord can be played. For this reason, only a limited number of chords were selected and programmed (mainly those chords that are open or basic chords). However, this indirectly places restrictions on the amount of variation in compositions that are generated. The results obtained in this thesis could further be improved by increasing the range of guitar chords that are used to generate new music. It would also be good if there was some sort of "global representation" that could be developed to "spell" the formation of a particular type of chord such as a major chord. This would be more of a challenge when the possible positions and the number of guitar strings used in a particular chord are considered.