>
TCTS Lab Staff
 
 

Arousal-driven synthesis of laughter

Examples of laughter phones

Example of laughter phones annotated in human laughs are available from this archive.

Per-frame arousal curves

Per-frame arousal curves of all the laughs from the AVLaughterCycle database can be downloaded from this archive. The files are simple text files with the computed arousal values. Instantaneous arousal is computed every 10ms (on 32ms windows): the first value in the file corresponds to a window centered at 16ms of the corresponding wav file, the second value is at time 26ms, the third value is at time 36ms, etc. Details about the algorithm are available in the paper "Arousal-driven Synthesis of Laughter" submitted to the Special Issue on Statistical Parametric Speech Synthesis of the IEEE Journal of Selected Topics in Signal Processing.

Examples of arousal-driven synthesis laughs

Hereunder are some example of synthesized laughs, after generation of phonetic transcriptions from per-frame arousal curves (methods G1-4), or directly from the corresponding human phonetic transcriptions (methods S3-S4). The corresponding human laugh is presented in the last columen (method H). Explanations about the different methods are available in the paper "Arousal-driven Synthesis of Laughter" submitted to the Special Issue on Statistical Parametric Speech Synthesis of the IEEE Journal of Selected Topics in Signal Processing.
G1 G2 G3 G4 S3 S4 H