
Table I. Performance of models on a set of piano music extracts. For the generic model, we have chosen the following set of parameters: H=10, A=1, =1, =0. | ||||
| Extract | Recording | M | Trained | Generic |
| Creation | Steinway | 2 | 100% | 100% |
| 3 | 90% | 84% | ||
| 4 | 85% | 78% | ||
| Moonlight | Elena Kuschnerovaa | 3–4 | 78% | |
| MIDI Synthesizedb | 3–4 | 91% | ||
| Variations | Andrew Koayc | 4 | 88% | |
| a Available from www.elenakuschnerova.com | ||||
| b Recorded using Winamp (www.winamp.com) | ||||
| c Available from http://music.download.com/ | ||||