Modern music information retrieval algorithms could be used to comprehensively characterize monaural recordings. However, there are currently no methods capable of analysing real-life binaural recordings in terms of their spatial audio content. The development of such analysers is important as they could be used for the search and retrieval of binaural recordings in terms of their spatial content. This paper presents a method for estimating one of the most important attributes pertinent to spatial audio scene description in music recordings, namely ‘ensemble width.’ The study builds on the concept of spatial analysis of binaural signals using ‘spatiograms,’ recently introduced by Arthi and Sreenivas . The method was developed using simulated anechoic conditions, whereby audio recordings were generated by the convolution of monaural recordings with head-related transfer functions (HRTF). According to the results obtained under the HRTF-dependent tests, the mean absolute error of ensemble width estimation was equal to 13.62°. Under the HRTF-independent test conditions, the error increased by only 1.99°, indicating that the generalization property of the method could be considered satisfactory.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.