art project using speech recognition
Printed From: MSSpeech-Forum
Category: Windows™ Speech Recognition Forums
Forum Name: New Users & General Questions
Forum Description: Ask questions, give and get answers.
URL: https://www.msspeech-forum.com/forum_posts.asp?TID=148
Printed Date: 15/Dec/2024 at 4:56pm
Software Version: Web Wiz Forums 12.02 - http://www.webwizforums.com
Topic: art project using speech recognition
Posted By: pietropalese
Subject: art project using speech recognition
Date Posted: 21/Dec/2009 at 4:00am
|
Dears I am working on an interactive installation but i dont have the knowledge of speech recognition programs This is how it should work : 1 microphone in every other room of the museum. in my room i will have for each microphone 1 1 pc with naturally speaking dragon, connected to each pc will be a printer. The printer will continuosly print out the conversations that will take place in the museum. I know this is pretty utopic but i dont need a perfect transcription of all the conversations, i accept errors. i just want to understand, before i buy the programm, if i will have or not a flow of text printed without doing any training to a particular voice in a space where there could be one or more person speaking. Do you think there could be a possibility to make my idea work? I thank you a lot in advance, Best, Daniel kemeny |
Replies:
Posted By: mmarkoe
Date Posted: 21/Dec/2009 at 6:43am
Impractical and would be a waste of money. Speech to Text speech recognition uses large speaker dependent vocabularies to achieve high accuracy. This means to work well (text produced is understandable), each user needs to train their individual speech profile. In addition, best results occur when speakers train themselves to enunciate each word and speak in phrases, not conversationally.
Read, http://www.mymsspeech.com/download/WSRGetStarted.pdf - "Getting Started with Windows Speech Recognition"
or
for Dragon NaturallySpeaking read, http://www.emicrophones.com/docDetails.asp?DocumentID=38 - "Key Steps to High Speech Recognition Accuracy"
and
Listen to the three sound files on http://www.emicrophones.com/articles/index.asp - How to Sound and How Not to Sound.
Marty
------------- Marty Markoe, MVP Microsoft Valued Partner See us at: http://www.mymsspeech.com |
pietropalese wrote:Posted By: pietropalese
Date Posted: 22/Dec/2009 at 5:31am
|
thanks for your fast answer! So from what i understand you think it is impossible to transcribe conversations? I have read on another web page something about speech analytic tchnology, that is used to transcribe telefon conversations in some callcenter software. Do you know if this is the same principle used in DNS? Could maybe this be a solution? Thanks a lot and marry chrismas, |
Posted By: mmarkoe
Date Posted: 22/Dec/2009 at 7:24am
Great minds make great strides. I think your project is impractical. Had you enough money, processing power and time it might be possible.
Why don't you contact http://www.nuance.com - Nuance (they manufacture Dragon) and ask.
Marty ------------- Marty Markoe, MVP Microsoft Valued Partner See us at: http://www.mymsspeech.com |
Posted By: pietropalese
Date Posted: 24/Dec/2009 at 7:08pm
|
Hi Marty I spoke with nuance but they dont seem to support me because the use of the programm i will do ist not tested so they dont know what to answer. I have got the program now, it works quite well also without traning and also transcribes conversations but not but the text seem then a bit absurd because of the errors and because you are not able to understand wich word belongs to wich person. But this is the nature of the medium so nothing to complain about. The problem now is the microphone. It work only it is place untl 20 cm of my mouth. I will try whit omnidirectional more sensibles omnidirectional microphones and see what distance i could manage to get. Best Daniel Daniel |
Posted By: mmarkoe
Date Posted: 24/Dec/2009 at 7:57pm
You cannot change the laws of physics. The further a microphone is from one's mouth, the stronger the pickup signal has to be because your voice drops off geometrically. Therefore the stronger pickup signal of an Omni directional microphone is also going to pick up stray sounds in your environment. This will make it anywhere from difficult to even more impossible for Dragon to recognize your speech as words. This is why close talk microphones work best for speech recognition. As a matter of fact, even the best noise canceling microphones work better in quiet environments than non noise canceling microphones.
Marty ------------- Marty Markoe, MVP Microsoft Valued Partner See us at: http://www.mymsspeech.com |