|A Cognitive Listening Sensor|
|Most sound recognition systems developed to date assume the incoming signal to consist of non-overlapping sounds with low levels of background noise. This is not the situation in which human hearing operates. The development of a noise robust verbal aggression detection system in 2007 was an important first step towards acoustic sensor technology that can be applied in real-life situations. The results in various applications showed promise, but also shortcomings.
In order to function in unpredictable environments with a wide variety of tasks, a sensor system needs to do more than just hear. It needs to listen. For that we need to know more about human listening, implement this knowledge in computational models and test these models against human performance in real situations. Only then can we hope to develop a listening sensor that can be called cognitive.
|Peter van Hengel, INCAS3|