Font Size: a A A

The Design And Realization Of Speech Recognizing Isolated Word Based On MPU.A

Posted on:2006-10-18Degree:MasterType:Thesis
Country:ChinaCandidate:H M ZhuFull Text:PDF
GTID:2168360152985388Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Considerable progress has been made in the field of automatic speech recognition (ASR)during the recent years. Now the technology of ASR is being put to practical uses. This subject is for the project of "The computer Animation based on visual fish model" in the research center .It is mainly focused on some technical difficulties in developing a subsystem of recognizing isolated word based on MPU, which implies the technology of speech recognition to 3D fish mimicking. The purpose of doing this is to make the already alive-like mimicking fish more splendid through speech controlling.In real applications, some factors such as new speaker, background noise and channel distortion degrade the performance of ASR system greatly. On the other hand, it's highly demanded for high-speed processor for speech signal processing and big storage resource for speech template that cost much. For degrading the burden ,it used semiconductor made by Philips and chose a subsystem of recognizing isolated word based on MPU.The realization of recognizing isolated word based on MPU comprise three parts of work including vertex testing, characteristics adoption and modeling matching .For avoiding loss data under the fast speech rate this paper provides theory for changing the strategy of traditional thinking of double buffer to trifle buffer. It is has been tested in practices that this technology indeed improves the reality in the system, avoids speech data losing under high burden. In the modeling matching, the matching between the sequence of template character and the sequence of speech character is implemented with dynamic timing-programming, such method make for the difference of different speech speed and can make out different word.
Keywords/Search Tags:speech recognition, Feature extraction, linear prediction, Dynamic Time-Warping(DTW)
PDF Full Text Request
Related items