The Echo files
The Brüel & Kjær Echo Speech Source Type 4720 is a small active speaker used for speech intelligiblity measurements where the voice spectrum, level and directivity are important. It is typically used in situations where no PA installation is used (f.i. in classrooms), or where the acoustics of a room or the background noise influence the speech intelligibility of a person speaking into a microphone. The Echo contains several preprogrammed stimulus signals that are recognized automatically by Dirac 6 when the Echo files are installed alongside Dirac. The Echo also has an output connector for playing these signals over a PA system. Here you can download the same stimulus signals as used by the Echo, and the setup file needed to have them recognized by Dirac. These signals can be played from an MP3 or CD player through a PA system to perform accurate speech intelligibility measurements.
Downloads
Installation
To use the Echo files you must at least download and install the Echo setup file and one or both sets of stimulus files. Note that the installation needs to be done on the same system on which Dirac 6 is installed and used for measurements. The stimulus signals can be written to an audio CD, copied to a smartphone or to any other device that can play them. The folowing signals are available:
| # | Name | Sequence [s] | Duration [m:s] | Description |
|---|---|---|---|---|
| 1 | intermittent | 10.92 | 10:00 | Intermittent pink shaped MLS signal |
| 2 | continuous | 10.92 | 10:00 | Continuous pink shaped MLS signal |
| 3 | voice | 10.92 | 10:00 | Male voice |
| 4 | pink BL | 60.00 | 01:00 | Bandlimited pink noise signal |
| 5 | pink | 60.00 | 01:00 | Pink noise signal |
| 6 | sine | 60.00 | 01:00 | 1 kHz sine at -3 dBFS |
| 7 | male MLS | 10.92 | 10:00 | MLS sequence with male speech spectrum |
Using the signals
Before starting a measurement the PA system's sound level needs to be set. The voice file (#3) can be used for this purpose. Play the voice file and adjust the PA system to a normal speech level. This is the sound level which would normally be used for speech. The speech level in the voice file is related to the level of the MLS signals, so once the PA system is set as desired using the voice file, the MLS files can be used for the actual measurements.
In Dirac, perform a level-calibration of your microphone or SLM and select the Echo measurement setup. No further options need to be set. Play the intermittent (#1) file over the PA system and start a measurement. The resulting window will show 2 channels. The top channel is the IR while the bottom channel contains the background noise.
Open the Speech table from the Parameters menu and verify that the INR values for all octave bands from 125 to 8 kHz are 20 dB or higher. If this is not the case then use pre-averaging to improve the INR. For the outer bands (125 Hz and 8 kHz) an INR somewhat lower than 20 dB may be acceptable because these bands do not weigh heavily in the final STI value. Particularly in the 125 Hz band an INR of 20 dB is not always easy to achieve due to background noise.
If pre-averaging is not enough to increase the INR to over 20 dB, you may be able to increase the PA system gain. Any gain increase of the PA system should also be entered as input gain on the Gain tab of the measurement window. In this case you should also consider whether the initial setup of the PA system was loud enough.
Verify that the impulse response fits within the 5 s window. If the impulse response (or the reverberation time) is much longer, you need to use the continuous MLS signal (#2). The background noise is not acquired automatically with this signal, so you will have to measure the background noise separately and enter these values in the measurement properties under the Speech tab. As a rule of thumb you should select an intermittent signal when SNR*RT < 120 dBs.
Note that you should not use the Echo files with a small speaker to simulate an actual mouth. Both the spectrum and level will be incorrect. Such a setup requires a careful equalization and calibration of the speaker, which has been done in the Echo speech source.
Dirac 4.x and 5.x
Earlier versions of Dirac cannot use the Echo files. However, the male mls file (#7) can be used with Dirac 4.x and 5.x to perform speech intelligibility measurements. Use the voice file (#3) to set the PA system's level as described above. Then use the male mls file (#7) to perform the measurement. Set the stimulus signal to external MLS in the measurement window. Depending on the amount of background noise and distortions in the PA system you may need to measure the signal and background noise values separately and enter the SNR values in the speech parameter window.