Detection of the amplitude modulation for modulating signals characterised by different crest factors

This study is concerned with the detection of amplitude
modulation (AM) of a tonal signal for modulating signals characterised by a
different crest factor and a low frequency. The purpose was to show that the
root-mean-square (RMS) value of the modulating signal is not a sufficient
measure of the modulation perceived by listeners. In the first experiment, using
the 2AFC method with an adaptation procedure, the AM detection thresholds were
determined for a 1-kHz tone modulated with 3-component harmonic complexes, the
components of which had the same amplitudes. The starting phases of the
complexes were chosen arbitrary to obtain the highest (2.44) and the lowest
(1.89) values of the crest factor of the same root-mean-square values. The
fundamental frequencies of the modulator were 1, 2 Hz and 4 Hz. The AM
thresholds gathered for these modulators were characterized by a certain scatter
that makes impossible the drawing of any conclusion concerning the influence of
the crest factor on those thresholds. Therefore in the second experiment
psychometric functions were determined for the detection of AM using the same
carrier and modulators. However, the results of this experiment did not provide
unambiguous evidence for a statistically significant effect of the crest factor
on the psychometric functions and the AM threshold values. Since the modulators
used in the first two experiments were characterised by too small differences in
the crest factor, the modulating signal in the third experiment was a harmonic
complex consisting of ten pure tones of the same amplitude. The starting phases
of all components were chosen so as to obtain signals characterised by the
highest (4.47) and the lowest (1.58) crest factor. For these modulating signals,
the probability of AM detection was found to depend significantly on the crest
factor, i.e. the detection of the AM was considerably easier when a modulating
waveform characterised by the highest crest factor was used. The results of this
experiment suggested that, at least for the low-frequency modulator, the
description of the AM thresholds by means of the RMS values of the modulator is
somehow misleading and does not express properly the abilities of the auditory
system to detect low-frequency amplitude changes. It has been also suggested
that for a modulating signal of a high crest factor the listeners can follow the
amplitude changes over time only when the changes are slow enough, i.e. when
sidebands produced by the amplitude modulation are not resolved in the auditory
periphery.