一种加权线性预测的声道面积估计方法
|
摘要
从语音信号声道滤波器中估计声道面积会受到声门波信号的干扰和非理想边界条件的影响。为了最小化这些因素的干扰,提出了一种加权线性预测估计声道面积的方法。首先采用相位倾斜的动态调节算法DYPSA(Dynamic Programming Projected Phase-Slope Algorithm)确定声门的开启和闭合点位置,然后利用加权线性预测方法计算声门闭相下声道模型的反射系数,最后根据声道滤波器和声管的等效模型,递推得到反射系数与声道面积的函数关系,并迭代求解声道面积。实验结果表明,计算同一段语音数据的声道面积,与核磁共振获得的标准声道面积比较,本方法估计的面积均方误差为0.03,声门波幅值在峰值一半以下的闭相法得到的均方误差为0.15,故本方法估计更为准确。
Existing methods that estimate the vocal tract area from vocal tract filters using speech signals suffer from inadequate elimination of the glottal wave,and the influence of non-ideal vocal tract boundary conditions. To minimize these effects on the vocal tract area estimation,we present a method that estimates the vocal tract area using closedphase attenuated weighted linear prediction. We use dypsa algorithm to estimate glottal closed instants and glottal open instants,then utilize weighted linear prediction with a specific attenuated main excitation trapezoidal segmentation weight function that attenuates the contribution of the glottal source to calculate the reflection coefficient of vocal tract model over the closed glottal phase. Finally we calculate the discrete vocal tract area function iteratively.The vocal tract area obtained by proposed method is more accurate than the reference method comparing with MRI data,and the result is that MSE_HPV = 0.15,MSE_WLP = 0.03.
Existing methods that estimate the vocal tract area from vocal tract filters using speech signals suffer from inadequate elimination of the glottal wave,and the influence of non-ideal vocal tract boundary conditions. To minimize these effects on the vocal tract area estimation,we present a method that estimates the vocal tract area using closedphase attenuated weighted linear prediction. We use dypsa algorithm to estimate glottal closed instants and glottal open instants,then utilize weighted linear prediction with a specific attenuated main excitation trapezoidal segmentation weight function that attenuates the contribution of the glottal source to calculate the reflection coefficient of vocal tract model over the closed glottal phase. Finally we calculate the discrete vocal tract area function iteratively.The vocal tract area obtained by proposed method is more accurate than the reference method comparing with MRI data,and the result is that MSE_HPV = 0.15,MSE_WLP = 0.03.
引文
[1]Sahoo S,Routray A.A Novel Method of Glottal Inverse Filtering[J].IEEE/ACM Tracsaction on Audio,Speech,and Language Processing,2016,24(7):1230-1241.
[2]Kinugasa H,Kamata H,Ishida Y.Spoken Word Recognition Using Vocal Tract Shapes[M].1993.
[3]Kestian A P,Smyth T.Real-Time Estimation of the Vocal Tract Shape for Musical Control[C]//Proceedings of the 7th Sound and Music Computing Conference,Barcelona,Spain.2010:206-211.
[4]Westbury J,Milenkovic P,Weismer G,et al.X-Ray Microbeam Speech Production Database[J].Journal of the Acoustical Society of America,1990,88(S1):S56.
[5]Denby,Stone M.Speech Synthesis by Real-Time Ultrasound Images of the Tongue[C]//Proc IEEE Int Conf Acoust,Speech,Signal Process,2004(1),685-688.
[6]Story B H,Titze I R,Hoffman E A.Vocal Tract Area Functions from Magnetic Resonance Imaging[J].Journal of the Acoustical Society of America,1996,100(1):537-554.
[7]Bresch E,Kim Y,Nayak K,et al.Seeing Speech:Capturing Vocal Tract Shaping Using Real-Time Magnetic Resonance Imaging[J].IEEE Signal Process Mag,2008,25(3):123-132.
[8]Mermelstein P.Determination of the Vocal-Tract Shape from Measured Formant Frequencies[J].Journal of the Acoustical Society of America,1967(41):1283-1294.
[9]Ladefoged P,Harshman R,Goldstein L,et al.Generating Vocal Tract Shapes from Formant Frequencies[J].Journal of the Acoustical Society of America,1978,64(4):1027-1035.
[10]俞振利,张礼和,曾尚璀.从语音信号的有限个共振峰频率估计声道面积参数的一个方法[J].电子学报,1997(7):32-37.
[11]Sondhi M M.Estimation of Vocal-Tract Areas:The Need for Acoustical Measurements[J].IEEE Transactions on Acoustics Speech and Signal Processing,1979,27(3):268-273.
[12]Wakita H.Direct Estimation of The Vocal Tract Shape by Inverse Filtering of Acoustic Speech Waveforms[J].IEEE Transactions on Audio and Electroacoustics,1973,21(5):417-427.
[13]Deng H,Ward R K,Beddoes M P,et al.Estimating Vocal-Tract Area Functions from Vowel Sound Signals over Closed Glottal Phases[C]//IEEE International Conference on Acoustics,Speech,and Signal Processing,2004:589-592.
[14]Schroeder M R.Determination of the Geometry of the Human Vocal Tract by Acoustic Measurements[J].Journal of the Acoustical Society of America,1967,41(4):1002-1010.
[15]Eye M,Infirmary E.Voice Disorders Database,version.1.03(cd-rom)[DB].Lincoln Park,NJ:Kay Elemetrics,1994.
[16]Wankhede N S,Shah M S.Investigation on Optimum Parameters for LPC Based Vocal Tract Shape Estimation[C]//International Conference on Emerging Trends in Communication,Control,Signal Processing and Computing Applications.2013:1-6.
[2]Kinugasa H,Kamata H,Ishida Y.Spoken Word Recognition Using Vocal Tract Shapes[M].1993.
[3]Kestian A P,Smyth T.Real-Time Estimation of the Vocal Tract Shape for Musical Control[C]//Proceedings of the 7th Sound and Music Computing Conference,Barcelona,Spain.2010:206-211.
[4]Westbury J,Milenkovic P,Weismer G,et al.X-Ray Microbeam Speech Production Database[J].Journal of the Acoustical Society of America,1990,88(S1):S56.
[5]Denby,Stone M.Speech Synthesis by Real-Time Ultrasound Images of the Tongue[C]//Proc IEEE Int Conf Acoust,Speech,Signal Process,2004(1),685-688.
[6]Story B H,Titze I R,Hoffman E A.Vocal Tract Area Functions from Magnetic Resonance Imaging[J].Journal of the Acoustical Society of America,1996,100(1):537-554.
[7]Bresch E,Kim Y,Nayak K,et al.Seeing Speech:Capturing Vocal Tract Shaping Using Real-Time Magnetic Resonance Imaging[J].IEEE Signal Process Mag,2008,25(3):123-132.
[8]Mermelstein P.Determination of the Vocal-Tract Shape from Measured Formant Frequencies[J].Journal of the Acoustical Society of America,1967(41):1283-1294.
[9]Ladefoged P,Harshman R,Goldstein L,et al.Generating Vocal Tract Shapes from Formant Frequencies[J].Journal of the Acoustical Society of America,1978,64(4):1027-1035.
[10]俞振利,张礼和,曾尚璀.从语音信号的有限个共振峰频率估计声道面积参数的一个方法[J].电子学报,1997(7):32-37.
[11]Sondhi M M.Estimation of Vocal-Tract Areas:The Need for Acoustical Measurements[J].IEEE Transactions on Acoustics Speech and Signal Processing,1979,27(3):268-273.
[12]Wakita H.Direct Estimation of The Vocal Tract Shape by Inverse Filtering of Acoustic Speech Waveforms[J].IEEE Transactions on Audio and Electroacoustics,1973,21(5):417-427.
[13]Deng H,Ward R K,Beddoes M P,et al.Estimating Vocal-Tract Area Functions from Vowel Sound Signals over Closed Glottal Phases[C]//IEEE International Conference on Acoustics,Speech,and Signal Processing,2004:589-592.
[14]Schroeder M R.Determination of the Geometry of the Human Vocal Tract by Acoustic Measurements[J].Journal of the Acoustical Society of America,1967,41(4):1002-1010.
[15]Eye M,Infirmary E.Voice Disorders Database,version.1.03(cd-rom)[DB].Lincoln Park,NJ:Kay Elemetrics,1994.
[16]Wankhede N S,Shah M S.Investigation on Optimum Parameters for LPC Based Vocal Tract Shape Estimation[C]//International Conference on Emerging Trends in Communication,Control,Signal Processing and Computing Applications.2013:1-6.