Parametric Factors Affecting Concatenative Sound Synthesis

Abstract

Many existing Concatenative Sound Synthesis (CSS) systems offer some form of user-controlled flexibility to its users. For example, users can select different audio features to be included as the basis of similarity between target and source sounds, or be provided with options to alter the pitch or loudness, or given the flexibility to set the similarity threshold between the target and the sound segments in the source database. However, with the exception of features selection, most of these control options are offered post-unit selection, i.e., after the segments are already selected and synthesised by the system. Post-unit selection transformation often means that re-selection of the sound segments to conform to the last minute adjustments entered by users. If these criteria were made clear before the selection of sound segments take place, it is possible that resulting sound will match the target more closely. This change will not only minimise the transformation needed, but also saves time as any ambiguities can be eliminated from the start. Thus, identifying the factors that affect synthesis result and including them in the system as options that users can control is the key to ensuring that the demands of users are communicated through to the system. This study investigated the effects of several parametric factors on the synthesis results on CSS. The significance of this study includes illustrating the need for increasing flexibility with respect to parametric options when developing a CSS system.