Method for executing automatic evaluation of transmission quality of audio signals using source/received-signal spectral covariance

• US 6,651,041 B1
• Filed: 02/09/2001
• Issued: 11/18/2003
• Est. Priority Date: 06/26/1998
• Status: Expired due to Fees

• Current Assignee: Ascom AG (TDC A/S)
• Original Assignee: Ascom AG (TDC A/S)
• Inventors: Juric, Pero
• Primary Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US09/720,373
• Time in Patent Office: 1,012 Days
• Field of Search: 704/206, 704/210, 704/228
• US Class Current: 704/228
• CPC Class Codes: G10L 25/69 for evaluating synthetic or decoded voice signals

Abstract

A source signal (e.g. a speech sample) is processed or transmitted by a speech coder 1 and converted into a reception signal (coded speech signal). The source and reception signals are separately subjected to preprocessing 2 and psychoacoustic modelling 3. This is followed by a distance calculation 4, which assesses the similarity of the signals. Lastly, an MOS calculation is carried out in order to obtain a result comparable with human evaluation. According to the invention, in order to assess the transmission quality a spectral similarity value is determined which is based on calculation of the covariance of the spectra of the source signal and reception signal and division of the covariance by the standard deviations of the two said spectra.

The method makes it possible to obtain an objective assessment (speech quality prediction) while taking the human auditory process into account.

Claims

1. Method for making a machine-aided assessment of the transmission quality of audio signals, in particular of speech signals, spectra of a source signal to be transmitted and of a transmitted reception signal being determined in a frequency domain, characterized in that, in order to assess the transmission quality, a spectral similarity value is determined by dividing the covariance of the spectra of the source signal and of the reception signal by the product of the standard deviations of the two spectra and is used in the calculation of transmission quality.

2. Method according to claim 1, characterized in that the spectral similarity value is weighted with a gain factor which, as a function of a ratio between the energies of the reception and source signals, reduces the similarity value to a greater extent when the energy of the reception signal is greater than the energy in the source signal than when the energy of the reception signal is lower than the energy in the source signal.

3. Method according to claim 2, characterized in that the gain factor reduces the similarity value as a function of the energy of the reception signal to a greater extent the higher the energy of the reception signal is.

4. Method according to one of claims 1 to 3, characterized in that inactive phases are extracted from the source and reception signals, and in that the spectral similarity value is determined only for the remaining active phases.

5. Method according to claim 4, characterized in that, for the inactive phases, a quality value is determined which, as a function of the energy Ep in the inactive phases, essentially has the following characteristic:

$A^{\frac{\log <br><br>\text{\hspace{1em}<br><br>}<br><br>10<br><br>\left(\mathrm{Epa}\right)}{\log <br><br>\text{\hspace{1em}<br><br>}<br><br>10<br><br>\left(E<br><br>\text{\hspace{1em}<br><br>}<br><br>\max \right)}}.$

6. Method according to claim 4, characterized in that the transmission quality is calculated by a weighted linear combination of the similarity value of the active phase and the quality value of the inactive phase.

7. Method according to claim 1, characterized in that before their transformation to the frequency domain, the source and reception signals are respectively divided into time frames in such a way that successive frames overlap to a substantial extent of up to 50%.

8. Method according to claim 7, characterized in that, in order to perform time masking, the spectrum of a frame has the attenuated spectrum of the preceding frame added to it in each case.

9. Method according to claim 8, characterized in that, before performing time masking, the components of the spectra are compressed by exponentiation with a value α

<

1.

10. Method according to claim 1, characterized in that the spectra of the source and reception signal are each convoluted with a frequency-asymmetric smearing function before determining the similarity value.

11. Method according to claim 10, characterized in that the components of the spectra are expanded by exponentiation with a value ε

>

1 before the convolution.