In this paper two common methods of HRTF/HRIR dataset interpolation, that is simple linear interpolation in the time and frequency domain, are assessed using a Normalized Mean Square Error metric. Frequency domain linear interpolation is shown to be the superior of the two methods, but both suffer from poor behavior and inconsistency over interpolated regions. An alternative interpolation approach based upon the Principal Component Analysis of the dataset is offered; the method uses a novel application of the Discrete Cosine Transform to obtain a functional representation of the PCA weight vectors that may be queried for any angle on a continuous scale. The PCA/DCT method is shown to perform favorably to the simple time domain method, even when applied to a dataset that has been heavily compressed during both the PCA and DCT analysis.
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.