Aerospace and Electronic Systems Magazine February 2018 - 55

Kruse
(The removal typically produces a distinct noise.) Is anything going on unnoticed in the cabinet or in the rest of the room, when the
music is playing, which might cover the sound of secret actions?
Using the recordings of the four microphones, some experiments have been done with beamforming algorithms. Again,
there are differences to standard applications, such as improving
noise-robustness for hands-free telecommunication and speech
acquisition [22], which has to deal with real-time constraints, but
often is limited to the steering of a far-field angular direction.
In the case considered here, the beam is steered towards certain
points of interest; unfortunately, in the area of the cabinet, the
cloth has some impact on sound propagation, so the free-field
assumption no longer holds. Rather than just suppressing noise,
various signals should be separated, e.g. the CD player, participants, and the audible phenomena in the center of the room. With
the setup of four microphones and a General Sidelobe Canceller
algorithm [23], some first results were obtained. For example,
the removal of the duct tape at the end of the session is clearly
discernible, such that the secret removal in the meantime seems
unlikely even when music is playing. Future measurements could
benefit from placing additional microphones at certain positions,
e.g. close to the CD player to more reliably separate the music
from other signals.
Yet another approach offered by microphone arrays is to measure the "spirit voices" from different positions and try to determine a direction of speech or a head orientation, based on the
direction-dependent damping of frequencies around the head [24]
(if there is a head...?).

Figure 5.

Distribution of formants F3 and F4.

Figure 6.

Comedian and his "kangaroo" voice.

can hardly be influenced intentionally and are very valuable for
speaker identification [17].
The formants were calculated with the Praat software [18] and
then exported to a separate software, which does some postprocessing (e.g. dealing with wrongly numbered/missing formants)
and calculates probability distributions of fundamental frequencies
and formants. Main interest has been put on the distributions of
F3 and F4. Figure 5 shows the results for three "spirit voices" and
the original voice of Warren Caylor: All voices have distinctively
different characteristics. Interestingly, "Tommy", who claims to
be a child, has the highest frequencies, which is in accord with a
shorter vocal tract. His F3 formant resembles the average of 8- to
10-year-olds, according to [19]. These profiles also remained the
same when the voices returned at further séances.
To illustrate that it is unlikely that a single person can produce
these different voices, the recording of a professional comedian
has also been analyzed while he is altering his voice for different
roles [20]. When he speaks the role of his "kangaroo" he sounds
very different, changing pitch and rhythm, but the formants F3 and
F4 remain the same (Figure 6).

BEAMFORMING AND FUTURE WORK
There are other aspects where audio signal processing helps investigating séance phenomena. Beamforming [21] allows focused
listening to what is going on in the room and to address typical
doubts: Does the duct tape, fixing the medium, remain in place?
FEBRUARY 2018

CONCLUSIONS
Mediumistic séances are maybe the most controversial topic in
the-anyhow much disputed-area of the "paranormal". But, believing or not believing should not be a requirement for doing serious research. Instead, the investigations have followed the simple
approach to collect as much data as possible under these admittedly unusual circumstances. The least thing to be said is: Audio
signal processing provides a powerful toolset, especially when our
vision-centered perception fails. Interestingly, even Warren Caylor, the medium, seemed to be very excited to literally see (in the
3D representations) what is going on in his séances.
Surprisingly enough-to say it with all care such a disputed
matter asks for: the common theories of trickery, such as helpers
among the participants, the medium getting rid of the tape and
producing the phenomena by himself, or the participants being
deceived into auditory illusions, were not supported by the measurements. Of course, this still does not need to be regarded as
compulsory proof for the "paranormal"; there might be other loopholes not taken into account so far. But maybe it is enough to attract some fellow researchers to this challenging puzzle?!

REFERENCES
[1]
[2]

IEEE A&E SYSTEMS MAGAZINE

Basel Psi Association. Switzerland, http://www.bpv.ch/, last access 1
May 2017.
Caylor, W. http://www.warrencaylor.co.uk/, last access 1 May 2017.

http://www.bpv.ch/ http://www.warrencaylor.co.uk/

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine February 2018