Application of psychophysical models for audibility prediction of technical signals in real-world background noise. (February 2015)

• Record Type: Journal Article
• Authors: Schell-Majoor, Lena; Rennies, Jan; Ewert, Stephan D.; Kollmeier, Birger
• Is Part Of: Applied acoustics. Volume 88(2015)
• Page Start: 44
• Page End: 51
• Publication Date: 2015-02
• Subjects: Psychoacoustics -- Auditory models -- Detection thresholds -- Masking effects -- Real-world signals -- Audibility; Acoustical engineering -- Periodicals; Periodicals; 620.2
• DOI: 10.1016/j.apacoust.2014.08.001
• Languages: English
• ISSNs: 0003-682X

Abstract

Abstract: A valid, objective computation of whether a real-world sound is detectable in a real-world acoustical environment is highly desirable in many noise control applications. However, most current prediction approaches have not been validated for this purpose and have not been tailored towards predicting the influence of certain signal features, such as the temporal structure or the spectral content of the masker or target. In order to evaluate the applicability of prediction approaches with respect to these signal features, detection thresholds of various real-world signals were measured for normal-hearing listeners. The detection thresholds depended on the temporal structure and spectrum of the target and the spectrum of the masker. The data were compared to predictions of five approaches ranging from time-averaged technical measures to psychoacoustic models, which incorporate these signal features to different extents. In general, the correspondence between predictions and the experimental data was better for the psychoacoustic models than for the results of the technical measures. Even though all models could account for most of the key effects in the experimental data, only the psychoacoustic models were able to predict the influence of the temporal structure of the signals. One of the models showed clear advantages in prediction performance, reaching an overall determination coefficient of R 2 = 0.94. This underlines the applicability of psychoacoustic models for correctly predicting audibility in real-world applications.