KnowBrainer Speech Recognition
Decrease font size
Increase font size
Topic Title: Training out Non-speech Sounds
Topic Summary: How to teach dragon to ignore non-speech noises
Created On: 12/20/2022 04:41 PM
Status: Post and Reply
Linear : Threading : Single : Branch
 Training out Non-speech Sounds   - Vittra_Whitetail - 12/20/2022 04:41 PM  
 Training out Non-speech Sounds   - Alan Cantor - 12/20/2022 05:50 PM  
 Training out Non-speech Sounds   - R. Wilke - 12/20/2022 11:19 PM  
 Training out Non-speech Sounds   - Alan Cantor - 12/21/2022 09:41 AM  
 Training out Non-speech Sounds   - Ag - 12/24/2022 10:07 PM  
 12/20/2022 04:41 PM
User is offline View Users Profile Print this message

Author Icon
New Member

Posts: 16
Joined: 06/15/2022

Checking with the community if there is an established way to train dragon to not detect certain sounds as speech. 

I have an issue with dragon hearing the sound of a ventilator as "six" and someone sniffing their nose as "it's". I have tried training these words independently so dragon knows what they do sound like but it's only partially effective. 


Any suggestions?

 12/20/2022 05:50 PM
User is offline View Users Profile Print this message

Alan Cantor
Top-Tier Member

Posts: 4466
Joined: 12/08/2007

I don't think you'll have luck training non-speech sounds. Approaches that might help, at least a little, include:

1. Reduce the influence of ventilator sounds by increasing the distance between the ventilator and the microphone.

2. Try moving the microphone from one side of the face to the other, so the head acts as a sonic insulator.

3. Try a better microphone and soundcard combination. Some combos do a better job of filtering out background noise than others.

4. Get good at quickly getting rid of unwanted words. The delete [word] command, specifically "delete it's" and "delete six", could become good friends!

 12/20/2022 11:19 PM
User is offline View Users Profile Print this message

Author Icon
R. Wilke
Top-Tier Member

Posts: 8018
Joined: 03/04/2007

Alan, while your suggestions 1-3 are very sound and I would only add one more by mentioning that is always considered best practice to turn off the microphone immediately when you're done dictating, I don't think deleting words such as "it's" or "six" from the vocabulary will do any good, because if you do, Dragon will pick up the ones sounding closest the next time.

It's like bad habits and the negative impact that they have; don't try teaching Dragon to ignore them, but avoid them, or better yet, eliminate them.


No need to buy if all you want to do is try ...

DragonCapture KB Download (Latest)
DragonCapture Homepage

 12/21/2022 09:41 AM
User is offline View Users Profile Print this message

Alan Cantor
Top-Tier Member

Posts: 4466
Joined: 12/08/2007

Hi RĂ¼diger,

I agree, deleting the words from the vocabulary would be a bad idea. My suggestion was to learn ways to get rid of unwanted words in the document. The "delete XYZ" and "delete XYZ through XYZ" commands do the job nicely in applications that support the commands. It's handy to be able to delete text without selecting and without moving the cursor.

The first skill I teach new Dragon users is to turn off the microphone immediately after speaking. Unfortunately, there's an exception: When using commands that enumerate matches, e.g., "delete XYZ," "copy XYZ," "bold XYZ" and so on, turning the microphone off dismisses the little numbers.

(Hmmm... Maybe I should toggle off the matching feature initially... From reading this forum, I realize some experienced Dragon uses dislike the feature, but I find it very helpful.)

 12/24/2022 10:07 PM
User is offline View Users Profile Print this message

Author Icon
Top-Tier Member

Posts: 967
Joined: 07/08/2019

I have had very good luck using the Nvidia Broadcast noise filtering software, as described in

It does require you to have a middle-end Nvidia GPU.

The post also mentions some other noise filtering software which I have not tried myself.


DPG15.6 (also DPI 15.3) + KB, Sennheiser MB Pro 1 UC ML, BTD 800 dongle, Windows 10 Pro, MS Surface Book 3, Intel Core i7-1065G7 CPU @ 1.3/1.5GHz (4 cores, 8 logical, GPU=NVIDIA Quadro RTX 3000 with Max-Q Design.

KnowBrainer Speech Recognition » Dragon Speech Recognition » Training out Non-speech Sounds

32472 users are registered to the KnowBrainer Speech Recognition forum.
There are currently 1 users logged in.
The most users ever online was 12124 on 09/09/2020 at 04:59 AM.
There are currently 322 guests browsing this forum, which makes a total of 323 users using this forum.

FuseTalk Standard Edition v4.0 - © 1999-2023 FuseTalk™ Inc. All rights reserved.