Data processing: speech signal processing – linguistics – language – Speech signal processing – For storage or transmission
Reexamination Certificate
1999-04-22
2001-05-15
Korzuch, William R. (Department: 2741)
Data processing: speech signal processing, linguistics, language
Speech signal processing
For storage or transmission
C704S217000
Reexamination Certificate
active
06233551
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method for measuring a voicing level used in a vocoder, and more particularly, to a method and an apparatus for determining multiband voicing levels using a frequency shifting method in a vocoder, which determines a voicing level based on autocorrelation.
2. Description of the Related Art
In general, a voice is represented by a pitch, a voicing level, and a vocal tract coefficient in a vocoder of low bit ratio. The pitch and the voicing level are modeled by an excite signal and the vocal tract coefficient is modeled by a transfer function. Here, the voicing level denotes a degree to which a voiced sound is included in a voice signal. The voicing level is one of the most important parameters for expressing a voice and plays a considerable role in determining the quality of the voice which passed through the vocoder. Therefore, a voicing level measuring method used for the vocoder has been constantly searched.
Traditionally, the voicing level simply determined the whole band to be voiced or unvoiced. This was employed in the LPC10:DoD 2.4 kbit/s standard vocoder. Dividing the voicing levels in two parts remarkably deteriorates the quality of the vocoder. Recently, a method in which the quality of sound is much improved is used. For example, in a multiband excitation (MBE) vocoder, the whole band is divided into a predetermined number of subbands in the frequency band of the voice and the respective subbands are determined to be voiced and unvoiced. Also, in a sinusoidal transform coder (STC), an analyze signal is expressed as a value between 0 and 1 by measuring periodical strengths of the analyze signal. According to the strengths, the band of the lowband frequency is determined to be voiced and the band of the highband frequency is determined to be unvoiced.
Methods of differently expressing the voiced levels in each subband are widely known.
First, there is the above-mentioned MBE vocoder method. In the MBE vocoder method, after normalizing the sum of the square of a difference between a synthesized spectrum obtained through modeling under the assumption that the whole band is voiced and an original spectrum, the normalized value is compared with previously set threshold values, thus determining whether the concerned band is voiced or unvoiced. Second, there is an STC method. While the MBE vocoder method determines the voicing levels on the spectrum, in the STC method, after normalizing the sum of the square between a synthesized periodical signal and an original signal in a time axis signal, the normalized value is compared with previously set thresholds, thus determining a voiced and unvoiced cut-off frequency. A spectral band less than the cut-off frequency and that more than the cut-off frequency are respectively determined to be voiced and unvoiced. In the above two methods, the voice levels are determined in each subband by comparing the difference between the original signal (or spectrum) and a synthesized signal (or spectrum) with the threshold value in a frequency or a time axis.
Third, there is an autocorrelation method of a time envelope signal. In this method, the voice signal is bandpass filtered for calculating a firm autocorrelation value in high frequency subband the time envelope of the filtered signal is estimated, and a normalized autocorrelation value is calculated from the estimated signal. The voicing levels of the respective spectral subbands are determined on the basis of the autocorrelation value. Fourth, there is an autocorrelation method of an upsampling signal. In this method, a time resolution is compensated by dividing the voice signal in each subband and performing upsampling with respect to the high frequency band. The normalized autocorrelation value is obtained from the upsampled signal and the voicing level is determined on the basis of the normalized autocorrelation value.
In the above two methods, the voicing levels are determined in each subband on the basis of the autocorrelation method. This is based on the fact that the autocorrelation value is larger as the voicing level of a voice is higher. Here, it is important how to calculate the autocorrelation value in the high frequency subband in which many errors are generated in calculating the autocorrelation value.
SUMMARY OF THE INVENTION
To solve the above problem, it is an objective of the present invention to provide a method for determining multi-band voicing levels using a frequency moving method in a vocoder for effectively obtaining an autocorrelation value in a high frequency subband and more firmly and effectively determining the voicing levels by obtaining the autocorrelation value after moving the frequency to an origin in each subband, on the basis of an autocorrelation method using the frequency moving method.
It is another objective of the present invention to provide an apparatus for determining multiband voicing levels for performing the above method.
Accordingly, to achieve the first objective, there is provided a method for determining voicing levels using a frequency moving method in a vocoder, comprising the steps of (a) applying a window to an input voice signal and obtaining a power spectrum from a voice spectrum obtained by Fourier converting a windowed signal, (b) moving the frequency of each subband to an origin after dividing the power spectrum into a predetermined number of subbands, (c) obtaining autocorrelation values of the respective subbands by inverse Fourier converting the power spectrum the frequency of which is moved to the origin, and (d) normalizing the respective autocorrelation values and determining the voicing levels of the subbands from the normalized autocorrelation values.
To achieve the second objective, there is provided an apparatus for determining voicing levels using a frequency moving method in a vocoder, comprising a band dividing portion for dividing a power spectrum obtained from a voice spectrum with respect to an input voice signal into a predetermined number of subbands, a frequency moving portion for moving the frequencies of the respective divided subbands to an origin, an inverse Fourier converting portion for obtaining autocorrelation values of the respective subbands by converting the power spectrum the frequency of which is moved to the origin by an improved inverse Fourier method of Goertzel, and a voicing level determining portion for normalizing the respective autocorrelation values and determining the voicing levels of the respective subbands from the normalized autocorrelation values.
REFERENCES:
patent: 5216747 (1993-06-01), Hardwick et al.
patent: 5583784 (1996-12-01), Kapust et al.
patent: 5809453 (1998-09-01), Hunt
patent: 5826222 (1998-10-01), Griffin
patent: 5890108 (1999-03-01), Yeldener
patent: 6023671 (2000-02-01), Iijima et al.
Cho et al., “A spectrally mixed excitation (SMX) vocoder with robust parameter determination,” Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Proceeding, vol. 2, May 1998, pp. 601 to 604.*
Beraldin et al., “Overflow analysis of a fixed-point implementation of the Goertzel algorithm,” IEEE Transactions on Circuits and System, vol. 36, Issue 2, Feb. 1989, pp. 322 to 324.*
Kim et al., “Use of spectral autocorrelation in spectral envelope linear prediction for speech recognition,” IEEE Transactions on Speech and Audio Proceeding, vol. 7, Issue 5, Sep. 1999, pp. 533 to 541.
Cho Yong-duk
Kim Moo-young
Burns Doane , Swecker, Mathis LLP
Korzuch William R.
Lerner Martin
Samsung Electronics Co,. Ltd.
LandOfFree
Method and apparatus for determining multiband voicing... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for determining multiband voicing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for determining multiband voicing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2529216