**Conceptual Overview**

We present a novel approach for modeling and synthesizing plucked-guitar tones by modeling a corpus of excitation signals using principal components analysis (PCA). PCA is applied to a corpus of excitation signals extracted from plucked-guitar recordings in order to develop a codebook of functions that can be used to model and synthesize the nuances of particular articulations. In particular, we analyze and model the excitation signals obtained by exciting the string using either a pick or finger while altering the dynamics of the attack.

Through analysis of the principal components, this technique offers insight into parameterizing the excitation signal, which has application for virtual music interfaces and can provide insight into the expressive intentions of the guitarist through audio analysis.

**Principal Components Analysis and Synthesis**

PCA determines basis functions and weights that can be used to describe any excitation signal in our corpus. Each excitation can be represented by a linear combination of the basis functions and weights.

Each excitation, **p**_{i}, is represented by a linear combination of basis vectors:

**p**_{i}: excitation signal**ṗ**: empirical mean of excitation signals*M*: length, in samples, of basis vectors**v**_{m}: mth basis vector*w*_{i,m}: weight for*m*^{th}basis vector

A codebook is developed for pick and finger articulations which contains *L ≪ M* basis vectors. The codebook entries are obtained from a weight histogram that determines the most essential basis vectors used in the excitation signals comprising the data set.

**Derived Excitation Signals and Principal Components**

Excitation signals obtained from plucking the string with a pick. | Excitation signals obtained from plucking the string with a finger. | Mean signal and first 3 principal components. |

**Sound Examples**

Description | Clip |

finger, piano to mezzo-forte to forte | |

finger, piano, scale | |

finger, mezzo-forte, scale | |

finger, forte, scale | |

pick, piano to mezzo-forte to forte | |

pick, piano, scale | |

pick, mezzo-forte, scale | |

pick, forte, scale |