DocumentCode :
3096288
Title :
Another approach for theoretical recognition method (II)
Author :
Sugiyama, Shigeki
Author_Institution :
Sch. of Eng., Univ. of Gifu, Gifu, Japan
fYear :
2010
fDate :
15-17 June 2010
Firstpage :
1724
Lastpage :
1729
Abstract :
We speak languages sometimes fast and sometimes slowly. We sometimes even stop speaking suddenly, and change the topic and start speaking about another subject. And also, we sometimes speak to a person whom we have met for the first time. Even in these situations, we can happily understand one another what we are talking about even though the speech signals wave forms are independent one another. But if we apply some of today´s speech analysis methods in these situations mentioned above, we may totally get different answers or even cannot apply those methods. And what is worse, application of those methods are quite difficult and it is not understandable what will be the causes in a practice situation. So on this paper, I have tried to study a simple method which can apply to in the difficult situations mentioned above by using Vector Analysis Method and Back Propagation Neural Networks Method, which mainly focuses on “Word feedback and Sentence feedback”.
Keywords :
backpropagation; natural language processing; neural nets; speech recognition; back propagation neural networks method; sentence feedback; speech analysis methods; speech signals waveforms; theoretical recognition method (II); vector analysis method; word feedback; Biological neural networks; Cities and towns; Feeds; Humans; Intelligent networks; Natural languages; Neurofeedback; Neurons; Speech analysis; Speech recognition; BPNN; fast method; sentence feedback; speech recognition; syllable feedback; word feedback;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Industrial Electronics and Applications (ICIEA), 2010 the 5th IEEE Conference on
Conference_Location :
Taichung
Print_ISBN :
978-1-4244-5045-9
Electronic_ISBN :
978-1-4244-5046-6
Type :
conf
DOI :
10.1109/ICIEA.2010.5515269
Filename :
5515269
Link To Document :
بازگشت