Section 5: Future Directions
Despite their inherent limitations, all of the machine testing methods that we’ve discussed can show good agreement if the system under test is reasonably well behaved.
But intelligibility testing is most consequential (and potentially most useful) when the system has problems severe enough to impair speech transmission. Such problems can arise from a variety of sources and conditions, many of which can “fool” any of the machine testing methods.
Contemporary sound systems are sophisticated complexes of diverse, interacting components. As the simplified diagram in Figure 6 illustrates, they invariably include signal processing elements whose effects on intelligibility, and on the instruments designed to measure it, may be difficult to predict.
While the consequences of relatively simple analog processing (such as equalization and limiting) generally are benign, the same may not be true of new, powerful digital signal processing technologies.
Figure 6 (click to enlarge)
For example, much attention is now focused upon using DSPs to “deconvolve” the response of a space in order to suppress echoes and subtract or add reverberation. Because the algorithms that are involved affect the time order of the signal, there may be large consequences if these devices are misadjusted. Furthermore, if speakers are repositioned, or the acoustics of the space changes (when a curtain is closed, for example), then the particular deconvolution likely will no longer be valid and may, in fact, cause very destructive effects.
None of the present machine measures for intelligibility accounts for time distortions. In fact, we could conceive of a hypothetical system that reversed the time aspect of a signal, like playing a tape backward: no machine method would show any decrease in the intelligibility score for such a system, though it would obviously render speech unintelligible.
What’s needed is an analyzer that’s sufficiently “smart” to detect all of the factors which affect intelligibility, and render a conclusive judgement, without relying heavily on the operator’s interpretation. But the unavoidable truth is that, as sophisticated as machine-based measurement systems may be, they cannot yet approach the complexity of the human ear/brain mechanism informed by a lifetime of experience decoding speech.
We can only model those aspects of that exquisitely fine-tuned mechanism that we have come to understand. The many remaining questions regarding how it works and what factors may affect it can only be answered by further research.
These papers were written by Ralph Jones, edited by Rachel Murray, P.E., and provided by Meyer Sound.
Go To: Section 1 Section 2 Section 3 Section 4 Section 5