VoxForge
VoxForge
NOTES:
This is a recording of Arthur the Rat with the 20db gain *disabled* on my sound card. I disabled the 20db gain in an attempt to limit "hissing" reported in my previous recording (jaiger-10212006.)
Due to this setting, these recordings are more quiet and I didn't bother trying to get my recording levels between 0.5 and 1.0 in Audacity. The 20db gain setting was what I needed to get my levels there in my previous recording.
I'm interested to hear if these are better or worse than my last recording.
-joe
Speaker Characteristics:
Gender: male;
Age range: adult;
Pronunciation dialect: General American English;
Recording Information:
Microphone: Logitech Precision PC Gaming Headset
Audio Card: intel8x0 chipset, builtin IBM Thinkpad T23
Audio Recording Software: Audacity 1.2.3
O/S: Debian GNU/Linux sarge
File Info:
File type: wav;
Sampling rate: 48kHz;
Sample rate format: 16bit;
Number of channels: 1;
PROMPTS:
ar001 Once there was a young rat named Arthur who never could make up his mind.
ar002 Whenever his friends asked him if he would like to go out with them,
ar003 he would only answer, "I don't know;" he wouldn't say yes or no either.
ar004 He would always shirk making a choice. His Aunt Helen said to him,
ar005 "Now look here! No one is going to care for you if you carry on
like this.ar006 You have no more mind than a blade of grass."
ar007 One rainy day the rats heard a great noise in the loft.
ar008 The pine rafters were all rotten, so that the barn was rather unsafe.
ar009 At last the joists gave way and fell to the ground.
ar010 The walls shook, and all the rats' hair stood on end with fear
and horror.ar011 "This won't do," said the captain; "I'll send out
scouts to search for a new home."
ar012 Within five hours the ten scouts came back and said,
ar013 "We found a stone house where there is room for us all.
ar014 There is a kindly horse named Nelly, a cow, a calf, and a garden with an elm tree."
ar015 The rats crawled out of their little houses and stood on the floor in a long line.
ar016 Just then the old rat saw Arthur. Stop. he ordered coarsely.
ar017 "You are coming, of course." "I'm not certain," said Arthur, undaunted,
ar018 "The roof may not come down yet."
ar019 "Well," said the old rat, "we can't wait for you to join us. Right about face! March!"
ar020 Arthur stood and watched them hurry away.
ar021 "I think I'll go tomorrow," he said calmly to himself, "but then again I don't know;
ar022 it's so nice and snug here,". That night there was a big crash.
ar023 In the foggy morning some men with some boys and girls rode up and looked at the barn.
ar024 One of them moved a board and saw a rat quite dead, half in and half out of his hole.
--- (Edited on 10/30/2006 14:45:37 [GMT-0500] by jaiger) ---
|
Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University. |
This recording sounds much better than the last one. The 'hiss' is much less pronounced, even at high volumes. So I did not have to use Audacity's noise removal tool.
I'm corresponding with someone with signal processing expertise (he develops audio codecs ...) and he says the strange noises that result from using Audacity's noise reduction algorithm (where the hiss in your recordings is replaced with 'space age' noises ...) is called 'musical noise'. He says it is difficult to get rid of. He has a noise removal algorithm that has much better results. But he recommends not using noise suppression at all - because any noise suppression algorithm always introduces a bit of distortion to the signal.
So, in your specific case at least, recording with the 20db gain *disabled* on your sound card is the way to go.
thanks,
Ken
--- (Edited on 10/31/2006 10:47 am [GMT-0500] by kmaclean) ---
|
Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University. |
Do you want me to re-record the last set of prompts with 20db gain disabled?
Will the project gain anything from a second recording of the same prompts from me or is my time better spent recording more unique prompts?
-joe
--- (Edited on 10/31/2006 13:10:55 [GMT-0500] by jaiger) ---
|
Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University. |
No, you don't need to re-record the last set of prompts with 20db
gain disabled. I used Audacity's noise reduction on it, which
would have reduced the audio quality some, but not enough to need a new
version of it - for now at least.
Yes, your time would be better spent recording other prompts.
BTW I have not created a newer release of the VoxForge acoustic model (like I said in a reply to your last post) because I am currently focusing on improving the whole speech submission system - the upgrade of the WebGUI content management system was part of this. I am working on getting the prompts to appear as tracks in Audacity, allowing users to save their recordings as a single Audacity project file, and upload a tar or zip file of that to the VoxForge web site.
thanks for you help with this project,
Ken
--- (Edited on 11/ 1/2006 11:48 pm [GMT-0500] by kmaclean) ---
|
Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University. |
