An Exploration of Virtual Auditory Shape Perception

[Previous Chapter][Table of Contents][Next Chapter]

6. Experiment Series

6.1. Apparatus and Interface

6.1.1. Spatialization

I used a Crystal River Convolvotron(TM) for acoustic spatialization. This device consists of a two card set that can be installed in Intel based IBM compatibles. It accepts up to four channels of audio input and produces a two channel (stereo) output. The computer to which I had access was a 33 MHz 80486DX. The Convolvotron(TM) was configured with the Universal Minimum Phase driver and an empirically derived head-related transfer function (HRTF) from data on the subject "SDO" of Wightman & Kistler [1989b]. This data set consists of 74 finite impulse response filters derived from recordings made with probe microphones in the ear canals of a human listener. White noise impulses were played from 74 spatial positions (12 azimuths, 6 elevations, and the top and bottom poles). In the minimum phase system, interaural phase information is added after other position related effects are calculated in order to avoid phase interpolation errors [Foster, et al., 1992].

6.1.2. Stimulus Presentation

For simplicity it would have been expedient to also use the 486 to generate the sound stimuli. However, for one of my planned experiments I would need four channels of output. Unfortunately, most inexpensive IBM compatible digital sound cards are limited to two channels. Furthermore (in spite of what some of the manuals say) it is not possible to install more than one of these cards. I was unable to obtain a PC audio card that would meet my requirements, so instead I used an expensive Macintosh 16-bit digital audio card: SampleCell from DigiDesign which the lab already owns. This meant using two computers instead of one, with two fans and two hard disks making copious quantities of noise. This also meant that I would need to establish a line of communication between the two machines.

The SampleCell card was installed in a Macintosh IIfx. The SampleCell card can store 16 MB of 16 bit digital audio and play four simultaneous channels at 44.1k samples/second (standard CD resolution). Sounds were remotely triggered by the `486 using MIDI (musical instrument digital interface) protocol. (See figure 6.1).

The headphones I used were Sennheiser model HD-250B. These headphones were selected because, in addition to their excellent linear frequency response characteristics, they have 16 dB of passive external noise attenuation.

Figure 6.1

Experimental Apparatus

Stimuli

The stimuli were similar to those used by Lakatos [1993a]. They consisted of 12-partial harmonic complexes with fundamental frequencies ranging from 1000-1150 Hz (see figure 6.2). These waveforms were synthesized using the TurboSynth which is sold by DigiDesign, the makers of the SampleCell card. TurboSynth allows synthesis by specification of individual frequency components. The characteristics of the stimulus were verified at the point just before they enter the Convolvotron using a LeCroy 9304 digital oscilloscope and performing a power spectrum transformation (again, see figure 6.2).

Figure 6.2

Power Spectrum of Stimulus With Fundamental Frequency of 1000 Hz

6.1.3. Interface

For these experiments I built the graphical user interfaces using Asymmetrix' Multimedia Toolbook software. Toolbook enabled me to easily build and modify the interfaces, and record subject data to disk files. With some additional work (and some additional help) I coerced Toolbook into controlling the Convolvotron(TM) and sending MIDI signals to the Mac IIfx to initiate stimuli. Many thanks to Brian Karr for writing the Windows Dynamic Link Libraries (DLLs) for millisecond timing, MIDI interaction, Convolvotron control, and for providing me with the templates I used to write the shape presentation DLLs.

The graphical interface was displayed on a 21 inch color monitor (Hitachi SuperScan Pro 21) at a resolution of 1280 X 1024 and in 256 colors.