Intel® Integrated Performance Primitives
Deliberate problems developing high-performance vision, signal, security, and storage applications.

DTMF tone detection issues

Emmanuel_W_
New Contributor I
214 Views
Hi,

I have been playing with the DTMF tone detection in the speech samples and I found several issues.
It seems to work fine for device with very clean input but not very well for others.

I have found the following issues.
- Missed detection
- Multiple detections
- Different results based on sampling rate.

I was able to make it work on a wide range of devices by doing the following modifications:
-The missed detection can be fixed by lowering the lower threshold from 7db to 6db
-The multiple detection is caused by the fact that the hold-on period is hardcoded to 320 which seems very short (devices I have tested with send a tone with duration between 80ms and 100ms). Devices which do not generate a clean signal will have multiple detections when some audio segement are not detected but previous 320 sample segment did trigger detection.
- Because the value of hold on is hardcoded to 320, result are different for 16khz and 8Khz.
- Overall I have increased the hold on period to be at least 50ms




0 Kudos
1 Reply
Vyacheslav_Baranniko
New Contributor II
214 Views

Hello,

Thank You very much for your valuable suggestions and comments. We will consider your input infurther development of this sample as soon as possible. Unfortunately, we are working in a resource constrains, so this may or may not be fixed in the nearest IPP version.

Anyway, nice to hear that you have succeeded in using the sample despite of its various shortcoming you have overcome yourself!

Regards,
Vyacheslav,
IPP speech coding

0 Kudos
Reply