Emotion Recognition from Disturbed Speech - Towards Affective Computing in Real-World In-Car Environments

Abstract:

Certain emotions can have a negative effect on the driver’s capability of safely operating the vehicle and can ultimately lead to accidents. Therefore, it would be beneficial if the vehicle was able to detect the emotional state of the driver and provide appropriate assistance to mitigate these effects. This study investiga- tes the influence of in-car acoustic characteristics and driving noises on emotion recognition from speech. The quality of the noisy speech samples was analyzed by calculation of SNR and CER[%]. Afterwards, classification experiments on high quality, in-car and noisy speech samples were carried out and evaluated. Data was recorded inside a car cabin in a simulator environment, resulting in realistic conditions where perturbations are being convoluted with the speech samples. For comparability with the state of the art, standard emotional speech databases were used for the evaluations conducted in this study. By considering the evaluated qua- lity and classification measures, we conclude that high quality emotional speech is most severely impaired in the car, and that highway noise reduces the performance of the emotion classifier strongly. This leads to further requirements for in-car emotion recognition.


Year: 2018
In session: Affective Speech
Pages: 208 to 215