Data processing: speech signal processing – linguistics – language – Modification of at least one characteristic of speech waves – Transformation of speech into a nonaudible representation,...
Patent
1997-01-13
1998-12-08
Knepper, David D.
Data processing: speech signal processing, linguistics, language
Modification of at least one characteristic of speech waves
Transformation of speech into a nonaudible representation,...
704233, H04M 322
Patent
active
058483845
DESCRIPTION:
BRIEF SUMMARY
FIELD OF THE INVENTION
This invention relates to the analysis of audio quality of a telecommunications system. It is desirable to monitor the performance of a telecommunications system in order to identify faults, and to ensure the optimal routing of network traffic.
DESCRIPTION OF RELATED ART
When specifying and monitoring performance of a telecommunications system it is highly desirable to apply a measurement which directly reflects those parameters which will determine likely customer satisfaction. By modelling the human senses, e.g. hearing, it is possible to estimate the subjective performance of real systems, e.g. the subjective audio quality. This approach to measurement, known as perceptual analysis, is an important emerging area.
Perceptual analysis techniques are important because of two main benefits: system, e.g. low-bit-rate speech coding schemes, which it is not feasible to measure with conventional engineering measurement methods. audio quality. This is exactly the information required to optimise system design and operation to provide the required subjective quality.
Existing perceptual analysis methods include both analogous and empirical models. Examples of analogous models are discussed in: Subjective Quality of Speech Coders", IEEE J. on Selected areas in Communications, Vol.10, No. 5, June 1992. on a Psychoacoustic Sound Representation", J Audio Eng. Soc., Vol. 40, No.12, Dec. 1992. Systems":, Procs. 10A, Vol. 13, Part 7, November 1991. Communications Systems Using a Speech-Like Test Stimulus", J Audio Eng. Soc., Vol. 41, No.12, Dec. 1993.
Empirical models are discussed in: on Attribute-Matching", Speech Comms, Early 1992. Normalization on NTIA Objective Voice Quality Assessment Method", DOC SQ-74.91, Dec. 1991. Transmission Quality in Telephone Networks", NTT Review, Vol.3, No. 5, September 1991.
All these models require knowledge of the input signal, i.e. the undistorted speech signal, in order to make comparative measurements from which the performance prediction is formulated.
FIG. 1 shows an idealised perceptual analysis system 8 receiving both the original version of a speech signal directly from an input 1 and the same signal distorted by the system 2 under test, from which it is able to estimate the subjectivity of the distortions added by the system under test.
The excitation in the human auditory system is algorithmically modelled for the original and distorted versions of the input signal, and a comparison of the outputs of the model for the two signal versions yields an estimate of the audible error. Further model stages interpret the audible error to estimate its subjectivity and hence the audio quality.
There is a further class of performance measurement which is of considerable commercial significance, but which cannot be addressed with existing perceptual analysis techniques. This requirement is to assess, non-intrusively, the subjective audio quality of network connections and routes carrying live traffic. Such a measurement would be highly advantageous in providing the following facilities: performance of the network to be assessed, action can be taken, and of particular connections to the "intelligent network" which will be able to route traffic accordingly. This will allow a more direct optimisation of subjective audio quality than is possible with conventional engineering measurement methods.
The performance of perceptual analysis for objectively assessing the subjective speech quality of complex systems has been demonstrated by Hollier M P, Hawksford M O, Guard D R, "Objective Perceptual Analysis: Comparing the Audible Performance of Data Reduction Systems", 96th AES Convention, Preprint 3797, February 1994. However, in order to apply perceptual analysis to the non-intrusive measurement, say of network traffic, knowledge of the input signal is required.
The input signal cannot be transmitted to the point of analysis in real time without being subject to the same kinds of distortion as are imposed by the system under test. With reference to FIG. 1, if the syste
REFERENCES:
patent: 4063031 (1977-12-01), Grunza
Hollier Michael Peter
Sheppard Philip John
British Telecommunications public limited company
Knepper David D.
Zintel Harold A.
LandOfFree
Analysis of audio quality using speech recognition and synthesis does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Analysis of audio quality using speech recognition and synthesis, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Analysis of audio quality using speech recognition and synthesis will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-190239