Laser Microphone Surveillance

2021 IEEE Conference of Russian Young Researchers in Electrical and Electronic Engineering (ElConRus) | 978-1-6654-0476-1/20/$31.00 ©2021 IEEE | DOI: 10.1109/ElConRus51938.2021.9396598

Sergey D. Koloydenko Ksenia V. Tcyguleva

Department of Computer Systems and Technologies Department of Computer Systems and Technologies
National Research Nuclear University MEPhI (Moscow National Research Nuclear University MEPhI (Moscow
Engineering Physics Institute) Engineering Physics Institute)
Moscow, Russian Federation Moscow, Russian Federation
[email protected] [email protected]

Abstract—This article describes the creation of an electronic FBI for example [5,6], find the use of it, because of its safety
device based on catching a laser beam, which can be used to and stealthiness.
eavesdrop from a safe distance. Photosensitive element receives
bounces of laser beam caused by micro-vibrations in response to With the development of technology, laser microphones
pressure waves created by noise. Voltage changes can be were upgraded: better receivers, amplifiers and different
withdrawn as sound after amplification. A laser beam can be auxiliary software were created. Overall construction became
directed into the room window, then it reflects off it and returns easier to use and produce not only for the government but also
to a receiver. This allows for safe and stealthy surveillance. But for private military companies. As components become
still, this technology can be upgraded. cheaper and more available, even a regular person can build a
spying device themselves nowadays, although it is forbidden
Keywords—laser; surveillance; photosensitive elements; sound by law [7]. The project has no intention of spying on other
capturing; speech to text conversion people and it was only used for tests on existing audio files.
I. INTRODUCTION Despite various advantages, technology has some
People always wanted to possess valuable information, so drawbacks: it is difficult to set up and to configure, analyzing
espionage history goes back into several centuries. With the raw audio can be hard. But still, there is a considerable amount
development of science, new technologies were discovered, of improvements which can be implemented.
and the government constantly wanted hi-tech devices in their The main objective of the research was to build a laser
arsenal [1]. Lasers became no exception. microphone and make it easier to perform surveillance with it.
Waves, produced by human voice vibrate nearby objects, Operating with text files is more comfortable than with raw
making it possible for an analog electronic device to convert audio, so the project was focused on converting speech into
these vibrations into the audio signal. It can be accomplished text. That required usage of computer software.
using “laser microphone”, which receives bounces of reflected
a laser beam off the vibrating object. Laser deflections caused
by vibrations can be transformed into audio signal, amplified,
and filtered [2].

Fig. 1. Simple depiction of laser microphone’s working method.

The technique of using a light beam as spying device was

invented by Leon Theremin in the Soviet Union in 1947 and
was used into “Buran” device, reliable on 500 m distance with
post-processed signal to remove noise[3], to spy on U.S,
British and French embassies in Moscow [4]. Laser
microphones were popular in the 90s but fell out of favor with
the rise of computer eavesdropping technologies which
provided more value. But now some anti-terror unis, CIA and
Fig. 2. Diagram for the device.

1991 978-1-6654-0476-1/21/$31.00 ©2021 IEEE

Authorized licensed use limited to: East Carolina University. Downloaded on June 21,2021 at 01:09:10 UTC from IEEE Xplore. Restrictions apply.
 II. RESEARCH MATERIALS AND METHODS
The whole device consists of two main parts: laser beam
and receiver. The idea of the project was to make a simple
device from cheap parts. All the components could be easily
obtained.
Laser in laser microphones is usually infrared because it is
invisible to the human eye, but it is hard to aim and should be
paired with red or any other color from the visible spectrum.
Using only red laser in tests is a viable option because photo
elements are sensitive enough for it.
On a receiving end there can be photoresistor, a simple
device that changes its resistance reacting to light, or
phototransistor, which has a faster response and acts as an
amplifier.
After being received, the sound should be amplified before
outputting because incoming vibrations are very small: even a
human ear can hardly detect them. Fig. 4. Test subject receiver’s circuit [8].

Fig. 5. Photoresistor from the night lamp and transistor.

Fig. 3. Red dot laser.

Before constructing the device, test subject was created to

prove that technology works. It consisted of 5mW 5 V 650 nm
red laser and receiver, based on cheap photoresistor from a
night lamp, made of cadmium sulfide. It can receive
wavelengths from 600 to 800 nm. NPN transistor was added to
the circuit to amplify the outcoming signal. The sound was
outputted on headphones, connected to the circuit. As a Fig. 6. Photoresistor’s sensitivity to wavelength graph.
replacement for window glass, speakers with attached glass
were used. The laser beam was aligned parallel along with Tests were successful: technology was proved. The output
receiver for best performance. sound of the test model was slightly hearable and recognizable,
but not comfortable enough for practical use even after
amplification, and the rate of noise was low.
The next step was creating a more sensitive device with
higher amplification with output on headphones.
The device consisted of 5mW 5 V 650 nm red laser, which
can be swapped to a 3mW 5 V 850 nm infrared laser for stealth
action, and receiver, based on a phototransistor (BPW85C). It

1992

Authorized licensed use limited to: East Carolina University. Downloaded on June 21,2021 at 01:09:10 UTC from IEEE Xplore. Restrictions apply.
can receive wavelengths from 620 to 980 nm. LM386 integral
amplifier was added to amplify the outcoming signal and 50
Ohm trimmer potentiometer to increase or decrease the volume
of an output signal. Later AUX cable was connected to the
circuit so it could be plugged into PC. The audio was output on
headphones connected to the PC.

Fig. 9. Phototransistor’s sensitivity to wavelength graph.

The sound was comfortably heard and recognized but the

noise was also loud. Overall, the performance of this device
was better than the test one and it was ready for the next step.
Although this device has many advantages: it is stealthy, it
is hard to track the position of an operator, it has a decent range
Fig. 7. Device receiver’s circuit [8]. of work. But there are disadvantages as well: it takes a lot of
time and patience to properly set up and to configure the
device. With that overall usage becomes not so easy. However,
some changes can be implemented to make it more
comfortable for an operator. For example, analyzing and
finding keywords in text is easier than in raw audio.
Conversion of speech to text assumed usage of neural
networks, so Python programming language was chosen for the
project. Python Kaldi library was the best option because it
provided an offline tool for speech recognition. Also, PyAudio
library was used to record audio and work with it. Before
starting the transformation, some changes can be applied to the
audio track. After amplification, some noises caused by other
vibrations became louder, so there was a common option to
remove them – using fast Fourier transform. In Python, FFT
can be performed with Numpy library. It relates a signal
sampled in time to the same signal sampled in frequency.
Fig. 8. Phototransistor, LM386 amplifier connected to resistors and
condensers.

This phototransistor reaches its peak of sensitivity during

the usage of an infrared laser. Using a red laser is still a viable
option. The laser must be slightly offset from the center of the
phototransistor or photoresistor so when it moves across the
surfaces, there is a corresponding change in voltage. The whole
device is very sensitive to ambient light so it should be
operated in darkness [9].

1993

Authorized licensed use limited to: East Carolina University. Downloaded on June 21,2021 at 01:09:10 UTC from IEEE Xplore. Restrictions apply.
Fig. 10. Power spectrum of frequencies.
Fig. 12. Removal of noisy frequencies after fast Fourier transform (blue –
This allows us to track spikes of most powerful frequencies signal before FFT, red – after).
which usually refer to clean signal and remove other
frequencies that are below some point [10]. After completing fast Fourier transform, parts of noise were
diminished or removed, as shown in (fig. 12), but one other
problem appeared, that did not allow to use this method.

Fig. 13. Appearance of additional noise after fast Fourier transform (blue -
signal before FFT, red – after).

The appearance of other noises led to a worse perception of

audio by a human ear and Kaldi recognizer, so overall using
fast Fourier transform was useless. Both codes with FFT and
without it could be found on GitHub [11].
The conversion was completed alongside with recording of
incoming audio. Kaldi did not always recognize words
properly and sometimes did not recognize them at all, but the
overall text is understandable: around 60-70% of words are
Fig. 11. Power spectrum of frequencies after removing noise. recognized correctly. It is possible to run the program several
times to form better results.

1994

Authorized licensed use limited to: East Carolina University. Downloaded on June 21,2021 at 01:09:10 UTC from IEEE Xplore. Restrictions apply.
 III. RESULTS REFERENCES
As a result, a simple laser microphone was built. Output [1] Volkman E. The History of Espionage: The Clandestine World of
audio was comfortably heard, however there were some noises. Surveillance, Spying and Intelligence, from Ancient Times to the Post-
9/11. World Carlton Publishing Group, 2008
Before converting speech to the text there was an option to
[2] Alhota G. Sound Waves Transmission via IR Laser: The Laser
remove excess noise using fast Fourier transform, but it the end Microphone. University of Benghazi, 2014
it only made it worse. Still, speech could be converted to text. [3] Galeyev B. M. Light and shadows of a great life: in commemoration of
Although speech recognition may not be the best, still some the one-hundredth anniversary of the birth of Leon Theremin, pioneer of
electronic art. Leonardo Music Journal, vol. 6, 1996, pp. 45-48.
keywords could be highlighted, so overall analysis of incoming
[4] Glinsky A. Theremin: Ether Music and Espionage. University of Illinois
information will be easier. Press, Urbana, IL, 2000
IV. DISCUSSION AND CONCLUSIONS [5] Wallace R., Melton К. H., Schlesinger H. R. Spycraft: The Secret
History of the CIA’s Spytechs, from Communism to Al-Qaeda.
The project showed that simple and cheap, but quite Dutton\Penguin Group, New York, NY, 2008
effective surveillance device can be built at home, and with [6] Lerner K. L., Wilmoth-Lerner В. Encyclopedia of Espionage,
software some additional utilities can be added. With better Intelligence and Security. Thomson-Gale, Detroit, United States of
components laser microphone becomes much more powerful, America, 2003
but this requires special equipment. [7] Petrochenkov S.D. Criminal liability for illegal circulation of special
equipment for surreptitious obtaining of information. Trudi Academii
As it is very easy to create a surveillance device for a upravleniya MVD Rossii [Works of Academy of administration of MIA
regular person, the question of security becomes important. of Russia], vol.1, 2013 pp. 108-110.
Even though rain, fog and light can counter laser microphone it [8] Lucid Science (2010). Available at: http://www.lucidscience.com/pro-
laser%20spy%20device-1.aspx (accessed 30 November 2020)
is still a solid device which is hard to track. Guaranteeing
[9] McDowell E. J., Ren J., Yang C. Fundamental sensitivity limit imposed
safety is very important nowadays because of the enormous by dark 1/f noise in the low optical signal detection regime. Optics
amounts of information people produce and work with. For Express, vol. 16, 2008, pp. 6822-6832.
example, using devices, which produce acoustic noise, or [10] Liu Q., Chen G., Liu X., Zhan J. Application of FFT and wavelet in
darkened windows. signal denoising. Journal of Data Acquisition & Processing, 2009
[11] GitHub (2008). Available at: https://github.com/Kolodez/PP (accessed
Even though the performance of speech recognition is good 30 November 2020)
it still can be upgraded, for example, working with dictionaries.
Trying to remove noise some other way is also an option. Still,
it is possible to make some other additions to the device.

1995

Authorized licensed use limited to: East Carolina University. Downloaded on June 21,2021 at 01:09:10 UTC from IEEE Xplore. Restrictions apply.