音声のピッチ同期線形予測分析のための改善法

Doctoral thesis English OPEN
劉, 麗清 (2015)
  • Publisher: Graduate School of Science and Engineering, Saitama University
  • Journal: 博士論文(埼玉大学大学院理工学研究科(博士後期課程)), volume 48, pages 55-58

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 Introduction 161.1 Overview of LP analysis . . . . . . . . . . . . . . . . . . . . . . . . 171.2 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 181.3 Dissertation Organization . . . . . . . . . . . . . . . . . . . . . . . 192 Pitch Synchronous Linear Prediction Analysis of High-PitchedSpeech Using Weighted Short-Time Energy Function 202.1 Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . 202.2 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.3 STE Function Based on Residual Signal . . . . . . . . . . . . . . . 222.4 Pitch Synchronous Analysis Based on Weighted STE Function . . . 252.5 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . 282.5.1 Results for Synthetic Speech Excited by Impulse Trains . . . 292.5.2 Results for Synthetic Speech Excited by Realistic Excitation 292.5.3 Results for Real Speech . . . . . . . . . . . . . . . . . . . . . 362.5.4 Stability of the Resulting All-Pole Filter . . . . . . . . . . . 392.6 Discussion and Summary . . . . . . . . . . . . . . . . . . . . . . . . 403 Noise Compensation LP Method Based on Pitch SynchronousAnalysis 423.1 Problem Description of Noise Compensation and Related Works . . 433.2 Noise Estimation Based on Pitch Synchronous Analysis . . . . . . . 453.2.1 Pitch Synchronization . . . . . . . . . . . . . . . . . . . . . 463.2.2 Enhanced Speech and Modified Noise Signals . . . . . . . . 473.2.3 Proposed Noise Estimator . . . . . . . . . . . . . . . . . . . 493.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . 523.3.1 Simulation for Verifying the Proposed Noise Estimator . . . 523.3.2 Results using Synthetic Speech . . . . . . . . . . . . . . . . 533.3.3 Results using Real Speech . . . . . . . . . . . . . . . . . . . 553.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614 Pink Noise Whitening Method for Noise Compensation LP MethodBased on Pitch Synchronous Analysis 624.1 Constraint Problem of Noise Compensation . . . . . . . . . . . . . 624.2 Proposed Prediction Whitening Filter . . . . . . . . . . . . . . . . . 634.3 Improved PSAS Method . . . . . . . . . . . . . . . . . . . . . . . . 694.4 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . 714.4.1 Results for Synthetic Speech . . . . . . . . . . . . . . . . . . 714.4.2 Results for Real Speech . . . . . . . . . . . . . . . . . . . . . 744.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 765 Linear Prediction Analysis of Crosscorrelation Sequence for VoicedSpeech 785.1 Problem Description of LP Analysis of Autocorrelation Sequence . . 795.2 LP Analysis of Crosscorrelation Sequence . . . . . . . . . . . . . . . 805.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . 825.3.1 Results for Synthetic Speech . . . . . . . . . . . . . . . . . . 825.3.2 Results for Real Vowel . . . . . . . . . . . . . . . . . . . . . 835.4 Pitch Synchronous LP Analysis of Crosscorrelation Sequence . . . . 855.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 906 Conclusions 916.1 Summary of the Work . . . . . . . . . . . . . . . . . . . . . . . . . 916.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92A Discussion of Noise Power When K is Odd 93
Share - Bookmark

  • Download from
    JAIRO via JAIRO (Doctoral thesis, 2015)
  • Cite this publication