July 24, 2010 at 15:31Scientists unsuccessfully try to synthesize laughter
Frequency patterns with strong harmonics make human laughter a rather complex subject for computer synthesis.
According to New Scientist, several groups of scientists are struggling with this task, trying to apply different approaches.
Jürgen Trouvain and colleagues from the University of Sarrbrücken (Germany), for example, are trying to simulate the movements of the vocal apparatus and air flow.
Gregory Beller (Grégory Beller) from the University of Pierre and Mary Curie (Paris), use a completely different approach. They took a regular text synthesizer and try to apply some distortion to the sound to create a simulation of various emotions , which are expressed in laughter.
At the same time, Shiva Sandaram(Shiva Sundaram) at the Deutsche Telekom research unit uses linear predictive coding to generate individual laugh elements (“ha”), as well as a simple timing algorithm.
Jerome Urbain from the University of Mons (Belgium) did find the easiest way: mixing and manipulating real-world samples of people.
Which approach is more successful is up to you. You can try to pass a kind of Turing test and try to determine the difference between human and synthesized laughter.
According to New Scientist, several groups of scientists are struggling with this task, trying to apply different approaches.
Jürgen Trouvain and colleagues from the University of Sarrbrücken (Germany), for example, are trying to simulate the movements of the vocal apparatus and air flow.
Gregory Beller (Grégory Beller) from the University of Pierre and Mary Curie (Paris), use a completely different approach. They took a regular text synthesizer and try to apply some distortion to the sound to create a simulation of various emotions , which are expressed in laughter.
At the same time, Shiva Sandaram(Shiva Sundaram) at the Deutsche Telekom research unit uses linear predictive coding to generate individual laugh elements (“ha”), as well as a simple timing algorithm.
Jerome Urbain from the University of Mons (Belgium) did find the easiest way: mixing and manipulating real-world samples of people.
Which approach is more successful is up to you. You can try to pass a kind of Turing test and try to determine the difference between human and synthesized laughter.