DNN-based laughter synthesis

Just like speech is part of our daily life, non verbal sounds are the same. Until nowadays speech has been taken into consideration by the most of researchers, whose based their research work on analyzing, recognizing, synthesizing and even translating it. Some of them decided to attack the non verbal sounds such as laugh, which is a means used by human to express his feelings and his state. So, during this paper, we will focus on analyzing and synthesizing it. For this purpose, deep neural network (DNN)and recurrent neural network (RNN) will be used as a benchmark of the synthesis procedure. This procedure consists of two main stages. The first stage is the analysis and preprocessing of the database and the second is the synthesis stage. We will perform an objective and subjective evaluation to ensure the performance of the DNN-based laughter synthesis.