Blog

The Echo Speech Source and ISO 3382-3

A draft version of ISO 3382-3 from 2009 contained the following paragraph:

NOTE Alternatively, for the STI measurements a directional source may be used. In accordance with IEC 60268-16 it should have directivity characteristics similar to those of the human head/mouth. The sound power spectrum should be as specified in Table 1. The main axis of the source should be pointed in the direction of the microphone.

In the final version this paragraph was removed, and the use of an omnidirectional source was mandated, but the sound power spectrum for a directional source remained in Table 1.

As is obvious from this change, there has been and can be discussion over the use of directional sources in ISO 3382-3. Occupants of an open plan office do not generally speak in all directions and in less diffuse environments the difference is measurable.

The main advantage of ISO 3382-3 is that it defines a common set of measurements and associated parameters that can be used to compare and classify spaces. Adding a directional source to this mix doesn't really help in this respect. So why was the option to measure ISO 3382-3 parameters with the Echo Speech Source added in DIRAC 6 build 5625?

The first, and probably most obvious reason is that current users of the Echo have requested this. Using the Echo is without a doubt the easiest and fastest way to measure the STI, and having a fully calibrated source renders a laborious system calibration unnecessary. Obviously the Echo will not produce results that are valid under the provisions of the ISO 3382-3 standard, but for a quick assesment of the situation it can save an enormous amount of time.

A second reason for supporting ISO 3382-3 parameters with the Echo is to help researchers investigate the effects of using a directional source versus an omnidirectional source. The Echo may also be helpful in developing new measurement techniques or even new standards. Along the same lines it is perhaps easier to find problematic reflections if a directional source is used. Again, we're not looking for standard compliant results, but for options to perform relevant measurements quickly and easily.

Note that there is another disadvantage to using the Echo, apart from not being in compliance with ISO 3382-3. The Echo is an asynchronous source and DIRAC therefore cannot derive the source-receiver distance from the measurements. These have to be entered manually, for instance using the 'Edit - Rotate - Set source-receiver distance' menu option.