Spatial Audio Processing: MPEG Surround and Other Applications

7.2 HRTF parameterization

The parameterization method that is outlined below employs several novel concepts. First, instead of processing existing HRTFs phase or magnitude spectra, a parameter-based analysis and synthesis approach is pursued using perceptually relevant transformation. Second, the proposed method enables modification of the amount of parameters to represent HRTFs in an adaptive fashion. Third, it is based on inter-aural phase relationships only, while completely discarding the absolute phase characteristic of HRTFs. Fourth, a comparison can be made between spectrally smooth and step-wise approximations.

7.2.1 HRTF analysis

The HRTF analysis step extracts parameters from an head-related impulse response (HRIR) pair of a specific spatial position. In a first step, the HRIR impulse responses h_l(n), h_r(n) are converted to frequency domain HRTFs using an M-point FFT, resulting in frequency-domain HRTFs H_l(m), H_r(m). Subsequently, the parameters are extracted that characterize the HRTFs by a set of perceptually motivated basis functions. The basis functions form a set of bandpass filters that mimic the known (spectral) limitations of the human auditory system. Given their bandpass characteristic, the basis functions are referred to as parameter bands. Each parameter band (or basis function) has an associated parameter band index b (b = 0,..., B − 1). The parameter band basis functions are specified in a matrix Q that has M rows and B columns. Each column (i.e. each parameter ...