We propose a statistical method for modeling and synthesizingsounds with both sinusoidal and attack transient components.In addition, the sinusoidal component can havepitch-changing characteristics. The method applies multivariatedecomposition techniques (such as independentcomponent analysis and principal component analysis) tolearn the intrinsic structures that characterize the soundsamples. Afterwards these structures are used to synthesizenew sounds which can be drawn from the distributionof the real original sound samples. Here we apply themethod to impact sounds and show that the method is ableto generate new samples that have the characteristic attacktransient of impact sounds.
|Title of host publication||Proceedings of Sound and Music Computing Conference|
|Publication status||Published - 1 Jan 2012|
|Event||Sound and Music Computing Conference - |
Duration: 1 Jan 2012 → …
|Conference||Sound and Music Computing Conference|
|Period||1/01/12 → …|