Table 1 Overview of the state-of-the-art research for guitar play recognition.
From: SpectroFusionNet a CNN approach utilizing spectrogram fusion for electric guitar play recognition
Refs. | Year | AI model | Classes | Performance measure in terms of accuracy (%) | Guitar database |
|---|---|---|---|---|---|
Christian Kehling et al.6 | 2014 | Novel algorithm | Plucking styles—fingerstyle, picked and muted Expression styles—bending, slide, vibrato, harmonics, and dead notes | 96% | Private dataset |
Li Su et al.7 | 2014 | SC + SG and SC + {CL, GDF, IFD} | Normal, hammer-on, pull-off, sliding, bending, vibrato, muting | 71.70% | Private Dataset |
Vincent Lostanlen et al.7 | 2018 | Scattering transform and supervised metric learning | 16 musical instruments with their playing techniques | 61.00% | Studio On Line (SOL) dataset |
Q. Xi7 | 2018 | Deep Salience multiple f0 estimation algorithm | JAMS file contains annotations such as tempo, key, instructed chords, performed chords and note level transcriptions | 46% | Guitar Set |
Marco Comunità et al.7 | 2021 | SetNet | 13 overdrive, distortion, and fuzz plugins. | 40.30% | Discrete Dataset |
MultiNet | 40.88% | ||||
FxNet + SetNetCond | 57.30% | ||||
FxNet(Monocontinuous dataset) | 90.09% | Continuous Dataset | |||
FxNet(Polycontinuous dataset) | 91.40% | ||||
Alexandros Mitsou7 | 2024 | SVM | Alternate picking, hammer-on, pull-off, slide, bend, vibrato, legato, tapping, sweep picking | 84.20% | Guitar style Dataset |
CNN | 81.10% |
