Realtime Software Synthesis for Psychoacoustic Experiments

New realtime sound synthesis software will allow psychoacoustic researchers to efficiently design and implement sophisticated test instruments that involve realtime interactivity with test subjects. Such interaction in psychoacoustic experiments has historically been constrained by the same limitations affecting realtime sound synthesis. A new model for sound synthesis, made possible through recent advances in computer hardware, supports software that synthesizes CD-quality audio in realtime and can base this synthesis on realtime user interactivity. We demonstrate this new software-based model in experimental settings, fully discuss its nature and abilities, and suggest general research applications.

Because of the large amount of information required to describe CD-quality sound and the time-sensitive nature of sound production, synthesis software for readily available computer systems has not been primarily designed to perform realtime sound synthesis (control of realtime synthesis based on user interactivity). Continuous data from subjects have not been easy to incorporate into current or subsequent stimuli. These limitations have made multiple pieces of equipment necessary in most setups. In the new realtime software synthesis model, all input, sound synthesis, and output are controlled by one device.

Several tools have been developed in the past to help manage audio and sound synthesis in a computer environment. One tool developed in an effort to greatly reduce necessary data flow is the Musical Instrument Digital Interface (MIDI). MIDI has several limitations, however, and because the sound wave is not described, different synthesizers will produce very different sounds when interpreting the same MIDI message, making it difficult to replicate studies done with MIDI.

Some software tools allow for elaborate descriptions of the sounds produced, but are not designed to do realtime synthesis based on continuous interaction. One piece of software, Csound, can do some realtime manipulation, but is not specifically designed for that purpose. Because of recent advances in processor power and standard memory configurations, realtime synthesis has become practical, and engineers are now designing for it. Three new examples of realtime synthesis software are SuperCollider, MSP, and Pd (Pure Data). These applications have the advantage of combining current processor power with a highly configurable computer interface, and can accomplish complex manipulation of sound in realtime. Each allows exacting control over almost all aspects of output, such as waveshape and frequency, based on user interaction. This provides for unprecedented flexibility in the description of the stimuli, the selection of the stimuli altogether, and the degree of precision in the calibration of responses. Researchers are also able to custom-design an instrument, and hear the instrument in realtime as they create it, greatly streamlining the development process.

Controlling sound synthesis in realtime is of great benefit to both the implementation and flexibility of the resultant experimental instrument. These software packages can have a great and immediate impact largely because of their intuitive interfaces, and relatively gentle learning curves. Any previous attempts at manipulation of sound based on realtime information would have required a great amount of programming on the part of the experimenter, and would have been limited by the technology available. The complexity of the synthesis algorithm determines the precision with which the sound can be controlled. More complex algorithms can greatly increase the number of computations required to produce one second of sound. The combination of powerful, relatively low-priced computers with this new software makes possible a degree of control and flexibility not previously available to most researchers.

MSP is a new extension to the MIDI-based Max environment, and allows the researcher to combine the strengths of both MIDI and realtime sound synthesis. Pd, written by Max's original author, has been designed from the ground up for realtime synthesis. Pd and MSP have graphics-based object-oriented programming approaches, as does Max. SuperCollider's object-oriented programming environment has a somewhat less graphical interface, but its easily configurable user interface facilitates the development of a convenient user environment with minimal programming effort.

While experimenters are given new tools with these packages, the presence of these instruments as software on a computer facilitates integration of standard techniques. For example, experiments could be inexpensively stored on writable CDs, which allow for random access of data, and the software could draw on a vast number of possible stimuli given the current state of interaction between the instrument and the subject. It would also be easy to record subjects' responses, and to access them randomly. For example, subjects could easily review their profile of previous stimuli, allowing for a more accurate relative description of a current stimulus. Experimenters familiar with MIDI could make use of a software implementation of MIDI, providing a virtual synthesizer within the computer.

Because of these new tools, researchers will have a new degree of flexibility and precision, enabling them to create more subtle and replicable test instruments that can interact with subjects in realtime.