1 / 4

17.0 Distributed Speech Recognition and Wireless Environment

17.0 Distributed Speech Recognition and Wireless Environment. References : 1. “Quantization of Cepstral Parameters for Speech Recognition over the World Wide Web”, IEEE Journal on Selected Areas in Communications, Jan 1999

Télécharger la présentation

17.0 Distributed Speech Recognition and Wireless Environment

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. 17.0 Distributed Speech Recognition and Wireless Environment References: 1. “Quantization of Cepstral Parameters for Speech Recognition over the World Wide Web”, IEEE Journal on Selected Areas in Communications, Jan 1999 2. “Exploiting Temporal Correlation of Speech for Error Robust and Bandwidth Flexible Distributed Speech Recognition,” IEEE Trans. on Audio, Speech and Language Processing, May 2007 3. “Robust speech recognition over mobile and IP networks in burst-like packet loss,” IEEE Trans. on Audio, Speech and Language Processing, Jan. 2006 4. “An Integrated Solution for Error Concealment in DSR Systems over Wireless Channels,” Interspeech 2006 5. “A Unified Probabilistic Approach to Error Concealment for Distributed Speech Recognition,” Interspeech 2005

  2. Client Input Speech Feature Vectors Linguistic Decoding and Search Algorithm Output Sentence Front-end Signal Processing Language Model Acoustic Model Training Speech Corpora Acoustic Models Language Model Construction Text Corpora Lexical Knowledge-base Grammar Lexicon Server Wireless Network Server Clients Distributed Speech Recognition (DSR) and Wireless Environment • An Example Partition of Speech Recognition Processes into Client/Sever • compressed and encoded feature parameters transmitted in packets • Client/Server Structure

  3. recognition results Wireless Network Speech Signal Server: Applications Feature Extraction Recognition Hand-held device-Client Speech Signal Wireless Network Feature Extraction Feature Compression Feature Recovery Recognition Hand-held device-Client Server Applications for voice communications (A) Feature Extraction Recognition Speech Signal Wireless Network Speech Decoder Speech Encoder Recovered speech Hand-held device-Client (B) Feature Extraction Recognition Applications Server Possible Models for Distributed Speech Recognition (DSR) under Wireless Environment • Problems with Wireless Networks • limited/dynamic bandwidth, low/time-varying bit rates • higher/time-varying error rates, random/bursty errors • Client-Only Model • speech recognition independent of wireless environment • limitation by computation requirements for hand-held device • Client-Server Model • proper division of computation requirements on client/server • bandwidth saving • not compatible to existing wireless voice communications • original speech can’t be recovered from MFCC compressed feature parameters • Server-Only Model • compatible to existing voice communications • seriously degraded recognition accuracy (A) • need to find recognition efficient feature parameters out of perceptually efficient feature parameters (B)

  4. ScalarQuantization V.Q. V.Q. V.Q. V.Q. V.Q. V.Q. (with optimized bit allocation) Client-Server Model • delta parameters evaluated at server • bit rate minimized • computation requirements minimized • recognition accuracy degradation minimized with acoustic models trained by quantized parameters (matched condition) • Split Vector Quantization for Feature Parameters (as an example) Sub-vectors • Error Protection • different error correction/detection schemes applied to V.Q. bit patterns for different parameters or different bits for scalar quantization • important bits well protected while extra bit rate for error protection minimized • correct identification of errors very helpful but at higher cost • compatibility with existing wireless networking platform needed • Error Concealment Examples • extrapolation • interpolation also possible • performed with those sub-vectors with errors (if identifiable) only xt: feature vector at time index t

More Related