Abstract
This work concentrates on validating a real-time perceptual model predicting distraction caused by audio-on-audio interference. The real-time model was recently developed on the basis of another successfully validated, perceptual distraction model, which is not able to calculate predictions in real time. Both models are non-blind, i.e., their inputs take target and interferer signals separately. This paper describes a validation experiment for the real-time distraction model, which compares the model's predictions to subjective distraction ratings obtained from a listening experiment. The accuracy of the real-time model is also compared to that of the original distraction model. The calculated root-mean-squared errors for a speech zone and a music zone were 10.2% and 12.6% for the real-time model, respectively, compared to 11.3% and 11.5% for the original model. The results indicate that the real-time model is able to predict the distraction with similar accuracy as the original model, and thus, is a suitable tool for sound-zone evaluation. Furthermore, the real-time capability of the model is considered to be vital for certain applications, including the evaluation of adaptive sound zones.
Highlights
Audio-on-audio interference is constantly present in our everyday lives, when two or more sound sources are competing for our attention
Where zj is the mean square of the jth gating block The features of the real-time distraction model are illustrated in Fig. 2 and described as follows: f10: maximum ITU-based loudness, when both the target and the interferer are active, f20: to-interferer ratio (TIR) based on the ITU loudness estimation, F03: calculated based on f20; see Eq (4), f40: the range of the ITU loudness estimation of the interferer signal at high frequencies [equivalent rectangular bandwidth (ERB) motivated bands 20–31], f50: percentage of temporal windows (400 ms, 25% overlap) where TIR based on the ITU loudness estimations is less than 13 dB
The results showed that the real-time model had a RMSE of 10.9% compared to the RMSE of 11.0% for the original model’s predictions of the same data (R€am€o et al, 2017a)
Summary
Audio-on-audio interference is constantly present in our everyday lives, when two or more sound sources are competing for our attention. The sound sources can be natural sources or they can be audio systems, such as a TV, a radio, or portable speakers connected to a smartphone This paper considers the latter case, where the sound sources are electrical devices, especially in the context of sound-zone systems. The model was originally trained by using a simple loudspeaker setup, consisting of only two loudspeakers, while at the same time considering that one of the main applications for the model would be the evaluation of sound-zone systems. This paper concentrates on validating the real-time distraction model by using a different sound-zone system, resulting in a different sound field, than during the development of the model.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
