In-Vehicle Speech and Noise Corpora - Digital Signal Processing for In-Vehicle Systems and Safety - page 147

Fig. 9.7 Illustration of in-class vs. out-of-class distances for each noise event in a car. Each dotted
region denotes a car and it encloses solid regions that denote session instances
Fig. 9.8 Inset and out-of-set distances for ACWC in three cars
The CU-Move corpus has been used extensively to understand the noise
properties in car environments and leveraging these properties for speech systems.
Examples of these studies include [ 3 , 4 ], and [ 5 ] by Akbacak and Hansen to
“environmental sniffing” of variations in environment to use most appropriate
models using the Rover scheme. In [ 6 ], the authors used the CU-Move corpus for
advancing voice-activated route navigation in car systems. Hansen [ 7 ] includes a
detailed description of the corpus along with the usage scenarios of CU-Move.