Detection and recognition of natural sounds

Human-machine interfaces and environment simulators increasingly depend on audio interfaces. Acoustic signals are used to provide warnings, feedback, information about the state of a system, and to enhance the immersive character of virtual reality environments. In order to decrease the mental workload of the listener, increase the speed of interaction, and minimize the chances for operational error, the audio signals (auditory icons) should have a natural character and clearly differ in their spatial, spectral, and temporal characteristics. Therefore, the design and selection of audio signals for specific applications should be based on the detectability and recognizability of the signals in the intended environments and on the meaningful connotations of the individual sounds. The present study was conducted to assess the detection and recognition thresholds of 30 pre-selected sounds and to determine the specific acoustic properties that make complex natural sounds effective auditory icons. The results of the study revealed a strong dependence of both types of threshold on the type of sound and a relative independence of both thresholds. The sound level difference between the detection and recognition thresholds varied from 1 to 13 dB and should be considered as an important criterion in auditory icon selection.