Evaluate localization performance with bandwidth filtered dirac stimulus
The suggestion is to evaluate statistically how the localization performance varies as a function of the bandwidth of the stimulus. The filters passbands may be suggest to discussion. For example, the filters could be based on the Bark scale.
"These bands have been directly measured in experiments on the threshold for complex sounds, on masking, on the perception of phase, and most often on the loudness of complex sounds. In all these phenomena, the critical band seems to play an important role. It must be pointed out that the measurements taken so far indicate that the critical bands have a certain width, but that their position on the frequency scale is not fixed; rather, the position can be changed continuously, perhaps by the ear itself."
Thus the important attribute of the Bark scale is the width of the critical band at any given frequency, not the exact values of the edges or centers of any band. [source]
Maybe we can try comparing the localization performance with Bark scales, and maybe we could try to vary the edges or centers of the bands in order to find if localization performance is more or less related to some specific bandwidths.
This work would require to
- apply a bandpass filter, which is already implemented in the scipy library. If using a dirac as base stimulus, it would only require the bandpass filter impulse response itself. These bandpass filters could also be obtained manually with some software, like Rephase, which can generate text files or wav files impulse responses.
- choose the appropriate length of the filtered impulses. Basically, the more steep the filters slope, and the lower the cut frequency, the greater length the filters should be. To check the appropriate length of the filters, the impulse response could be generated as a .wav file and then imported in Room Eq Wizard, which is a software than can plot the impulse response in many different and useful representations (impulse reponse in frequency domain, in temporal domain, analysis of reverberation time, clarity indices, spectrogram etc.).