Question 13/XII - Methods for the evaluation of non-linear distortions
(continuation of Question 13/XII studied in 19851988; new wording)

Non-linear distortion, in its most general sense, occurs in any system
where the output is not linearly related to the input. Currently, there
are several different methods for objectively quantifying non-linear
distortion some of which are only applicable with a particular input
signal (e.g. harmonic distortion). The following Questions should
therefore be studied:

1. What objective methods are available for evaluating the subjective
effect of non-linear distortion on speech transmission quality?

2. What are the applications/limitations of these methods, in terms of
predicting the subjective perception of non-linear distortion and can a
single method be recommended?

3. What test signals are appropriate for these methods?
(Attention is drawn to Recommendation P.50 which specifies a CCITT
artificial voice and to Question AA/XII)?

The results of the study should be applicable to the testing of
individual pieces of equipment, or in conjunction with models being
developed under Question 7/XII, as a more general tool for use in
transmission planning of complete connection.

Notes - Points to consider when evaluating the various methods for
measuring non-linear distortion include:

1. Correlation with subjectively derived results for a wide variety of
non-linear processes, including digital encoding schemes.

2. Capability of operating on actual hardware. This implies that the
method must be able to take into account any time delay existing between
input and output samples (at least up to delays typically encountered).

3. Capability of including other network impairments (e.g. frequency
response, circuit noise absolute listening level etc.) in the
calculation model.
ANNEX

(to Question 13/XII)

During the 1985-1988 study period, three different objective methods for
evaluating non-linear distortion were proposed. Since it was not
possible to gain agreement on a single method, each of these methods has
been retained for further study. The methods are:

1) A method from BNR based on the Coherence Function

(See COM XII-46 "Objective evaluation of non-linear distortion effects
on voice transmission quality" and COM XII-175 "Re-evaluation of the
objective method for measurement of non- linear distortion")

2) A method from NTT based on the Cepstral Distance

(See COM XII-7 "Proposal of study items on objective quality evaluation
method for voiceband codecs", COM XII-8 "Proposal of objective quality
measure for voiceband codecs" and COM XII-86 "Objective speech quality
estimation of non-linear telecommunication devices by LPC Cepstrum
distance measure")

3) A method from France based on the Information Index (See COM XII221
"Calculation of transmission performance from objective measurements by
the information index method").