JOURNAL DE PHYSIQUE IV

Colloque C5, supplCment au Journal de Physique 111, Volume 4,mai 1994

Underwater voice communications using digital techniques

B. WOODWARD and H. SARI

Department of Electronic and Electrical Engineering, Loughborough University of Technology,

Leicestershire LEll3TU; UK.

Abstract: A digital signal processor (DSP) based voice communication system for divers with
a 2400 bitls transmission rate is described. Each speech signal frame is analysed using linear
predictive coding (LPC) and the essential parameters such as amplitude, pitch period and digital
filter coefficients are extracted. These parameters are transmitted through a water channel at
a rate of 2400 bitls by employing digital pulse position.modulation (DPPM).

1. INTRODUCTION

The role of communications in diving is important not only from the point of view of information
exchange and operational efficiency, but also to achieve greater safety. Since 1905, divers have been
using voice communication systems, the early versions having a direct wire link between a hard hat diver
on the bottom and a supervisor at the surface. Like the diving equipment of the time, these systems were
simple and unsophisticated, and they were replaced in the 1960s by through-water acoustic systems which
were more suitable for divers using Self Contained Underwater Breathing Apparatus (SCUBA).

There are several commercially available analogue wireless diver communication systems [I-31. Some
early systems used baseband communication methods in which acoustic signals were transmitted without
modulation. However, these systems were affected by ambient acoustic noise which is dominaqt below
6 kHz [4]. In most subsequent versions this problem has been obviated by shifting the spectrum of the
baseband signals to a higher band and transmitting by using a modulation method such as amplitude
modulation (AM), double sideband suppressed carrier (DSBSB), single sideband (SSB) or frequency
modulation (FM). In view of its lower power consumption, SSB has been the most commonly employed
technique [S]. When used underwater with acoustic signals for diver-to-diver or diver-to-surface
communication, SSB has been shown to provide all the proven advantages of its counterpart for radio
transmission. Nearly all the analogue systems that are presently on the market are virtually identical in
their technical specifications and have not progressed beyond the technology of three decades ago. It is
all too clear that the rapid development in digital signal processor technology and the advent of mobile
communication systems has been completely ignored by the manufacturers of divers' voice
communication systems.

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jp4:1994598

C5-470 JOURNAL DE PHYSIQUE IV

2. FEASIBILITY OF DIGITAL SPEECH CODING FOR UNDERWATER COMMUNICATIONS

In this research project speech data is transmitted by an acoustic transducer with a resonance frequency
of 70 kHz, therefore a suitable kind of speech coding method has to be applied. After sampling the
speech signal at a rate of 8 kHz,all the coded information representing this sample must be transmitted
during the corresponding sampling period of 125 ps. There are several wavefom quantization and
encoding techniques for speech signals that have been widely used in telecommunications since the 1950s.
In particular, we consider adaptive differential pulse code modulation (ADPCM).

In 1990, the Consultative Committee for International Telephone and Telegraph (CCITT) approved
Recommendation G.727 for embedded ADPCM [6]. This specifies that the ADPCM algorithms can
operate with 5, 4, 3 or 2 bits per sample, i.e. at rates of 40, 32, 24, and 16 kbitls. Because of the
limitation of the projector, with a resonant frequency of f0=70 kHz, only 16 kbitls, or 4 cycles per bit,
may be used for speech transmission. Therefore, the maximum time allowed to transmit each bit is less
than 62,Sps. In a noiseless channel, this rate is acceptable. But during transmission through water, the
acoustic signal is seriously distorted by multipath propagation, which makes signal recovery very
difficult. A method requiring a lower transmission rate must therefore be employed.

One possibility is to use one of the so-called parametric coding methods, which are recommended for
low bit rate coding of speech signals [7, 81. These methods reduce the transmission rate to 400 bitls by
separating the excitation component of the speech from the spectral envelope component. The excitation
is then characterized as either a pulse train for voiced sounds or noise for unvoiced sounds. The spectral
envelope can be characterized by the parameters of a digital filter having the same transfer characteristics
as the vocal tract.

Linear predictive coding (LPC) provides an accurate representation of the relevant speech parameters that
can reduce transmission rates at the expense of a reduction in the quality of the reproduced speech [9].
In an LPC voice coder, efficient speech synthesis is achieved by transmitting frames of the speech
waveform as a set of parameters, which are (i) the amplitude, (ii) a voiced/unvoiced decision and pitch
period for voiced sounds, and (iii) the filter coefficients.

In the linear prediction representation, the speech signal is modelled as the output of an all-pole filter
H(z).
H(z) = G
P
I + a [klz - k
k=l

where p is the number of modelled poles, G is the gain of the filter and the a[k]s are the coefficients
characterizing the filter. Generation of the synthetic speech sequence requires a knowledge of the pitch
period, the linear prediction coefficients and the power of the waveform in each speech frame. These
parameters are computed by means of a TMS320C30 digital signal processor (DSP) [lo]. The advantage
of the LPC method is that the speech transmission rate is decreased to 2400 bitls, including error
correction codes. In comparing the wavefom and parametric coding methods, as can be seen from their
transmission rates, the LPC method is superior to ADPCM and is therefore preferred for underwater
acoustic transmission.
3. PULSE POSITION MODULATION FOR UNDERWATER SPEECH TRANSMISSION

Digital modulation techniques, such as amplitude shift keying (ASK) I1 11, frequency shift keying (FSK)
[12] and phase shift keying (PSK) [13] can be employed for underwater acoustic data transmission. As
an alternative, pulse position modulation (PPM) is also an effective technique. The analogue PPM
method is suited to serial data transmission systems [14]; its mode of operation is to sample the analogue
signal and to transmit a short pulse that is delayed with respect to the sampling instant by an amount that
is a linear function of the signal amplitude. The drawback is that during transmission a pulse can be
seriously distorted by multipath propagation. Therefore estimating its relative position at the receiver
is difficult, especially for high data rates.

Since the encoded speech signal is digital in nature, a digital pulse position modulation (DPPM) system
is proposed. Because the speech parameters are digitally coded with a transmission rate of 2400 bitls
each 1-second of data is divided into frames consisting of 3 bits, i.e. 800 frames per second as shown
in Fig. 1. A frame, with a period of T, = 312400 s (1.25 ms), is further divided into 10 slots, each of
width T, = 125 ps; these comprise 8 data slots which represent the eight possible 3-bit combinations of
data and two guard band slots. A single PPM pulse is transmitted in each frame, allocated to one of the
slots according to the digital information. The demodulation of such signals therefore requires two clock
signals, one at the frame frequency of 800 Hz and the other at the slot frequency of 8 kHz.

frame 1 frame 2 frame 3

I 1

000 001 010 011 100 101 110 111 GUARD BAND

Fig. 1. Digital PPM signal format for 8 data slots per sub-frame

Different synchronization schemes have been suggested in the literature for digital PPM transmission
[15]. The system described here operates asynchronously, since there is no phase synchronism between
the transmitter of one diver's system and the receiver of another. But in each separate system the
transmitter and receiver are controlled by the same clock, which is simply a multiple of the slot
frequency. Detection of the digital pulse position signal is achieved with the system shown in Fig.2.

Since a digital pulse representing the position of the data modulates a 70 kHz sinusoidal waveform, a
narrow bandpass filter is included in the receiver. The output of this filter is then passed through an
envelope detector. Consider the situation when one diver starts speaking. The filter output of another
diver's receiver first detects eight synchronization data pulses, i.e. 1111 111I,, which trigger a 4-bit
counter running 8 kHz. The outputs of the counter control the integrators. When it counts to 1000, the
outputs of the eight integrators are compared to a threshold level set just above a level representing the
amplitude of the unwanted multipath signal. During the pulse transmission interval the output of a
particular integrator will be high and this becomes the input to a 8-to-3 decoder. The 3-bit decoder
output is the transmitted parameter that is fed to the DSP-based LPC synthesizer.
 JOURNAL DE PHYSIQUE IV

INTEGRATORS COMPARATORS

11
COUNTER

CONTROL
COUNTER

Fig.2 Analogue system for digital pulse position detection.

REFERENCES

Overfield T., Underwater Systems Design, (MarchIApril 1988) pp.8-13.

Clark A., Underwater Systems Design, (January 1989) pp. 16-18.
Peck M.J., Sea Technology, (1992) pp. 6 1-65.
Virr L.E., ZEE Proc 6 (1987) pp. 547-576.
Woodward B., "Underwater Telephony: Past, Present and Future," Colloque De Physique 2
(1990) pp. C2-591-C2-594.
Sherif M.H., Bowker D. O., Bertocci G., Orford B. A. and Mariano G. A., ZEEE Trans. Comm.,
2 (1993) pp. 391-399.
Rabiner L.R. and Schafer R.W., Digital Processing of Speech Signals (Prentice-Hall, New Jersey
1978).
Deller Jr. J.R., Proakis J.G. and Hansen J.H.L., Discrete-Time Processing of Speech Signals
(MacMillan Publishing Comp., New York 1993).
Makhoul J., Proc ZEEE 63 (1975) pp.561-580.
Tremain T.E., Speech Technology 1 (1982), pp. 40-49.
Dawoud M.M., Halawani T.U. and Abdul-jauwad S.H. Znt. J. Electronics 72 (1992) pp. 183-196.
Brock D.C., Bateman S.C. and Woodward B., Ultrasonics, 24 (1986) pp. 183-188.
Falahati A., Bateman S.C. and Woodward B., ZEEE J. Ocean. Eng., OE-16 (1991) pp. 12-18.
Riter S. and Boatrigth P.A., "Design considerations for a pulse position modulation underwater
acoustic communication system," Digest IEEE C o d . Engineering in the Ocean Environment,
Panama City, Fla., September 1970 pp. 21-24.
Ling G. and Cagliardi R.M., ZEEE Trans. Comm. COM-34 (1986) pp. 1202-1208.