|
INTRODUCTION
This paper will describe the use of digital signal processing as it relates to audio surveillance
applications. Digital Signal Processing, DSP, processors are microprocessors that are specialized for
a specific task by programming classes of algorithms.
WHAT IS DSP?
The first successful, commercial, DSP processor was introduced in 1979 by
Intel Corporation. Before this analog signal processing existed, and is
still used today by many agencies. In either approach (analog or digital)
to the processing of signals, the desired result is to modify an input
signal to obtain a more meaningful output signal. The output signal may be
an audio intercepted coversation with noise from a radio or television
removed from the background. An example of analog signal processing is
shown in Figure 1. This figure could represent the filtering of a body-wire
radio signal that has been distorted while in transmission.
The problem with analog signal processing, for surveillance applications, is
that it is not possible to manufacture an ideal filter. Temperature,
power-supply variations, and component accuracy introduce errors into the
process. This coupled with the fact that an exact copy of stored analog
signals are impossible to obtain; results in analog signals being the less
desirable method of signal processing. A conversation on a cassette tape is
an example of stored analog signals. Upon play back from the cassette a
certain level of noise is introduced by reading the signals.
Digital signal processing is the digitizing of analog signals sampled at
regular intervals. Digitizing is the coversion of an analog signal to a
binary format. When the signal is sampled, at fast rate, a highly accurate
representation of the signal is acheived. Illustrated in Figure 2a is an
analog signal. The same signal in a digitized form is given in Figure 2b.
There are many advantages to using DSP signal processing. One of the most
important advantages, from a hardware point of view, is the reduced number
of components required to perform the desired processing. DSP design offers
the advantage of programmability. It is now possible to perform operations
that are difficult or simply not possible in the analog world, such as,
correlation and decorrelation with a time delay.
APPLICATIONS
Digital-Audiotape Technology (DAT), Compact Disc (CD), adaptive noise
cancellation processors, and semi-conductor recorders all use digital
signal processing. In fact any analog circuit can be replaced by DSP.
DSP FOR AUDIO RECORDERS
The analog tape recorder for many years has remained the method most widely
used for surveillance operations. Digital technology has made possible the
introduction of the DAT and semi-conductor recorders. DAT recorders store
signals on a cassette that resemble an 8mm video tape; because of this
they are not useful in covert body worn applications. The semi-conductor
recorder does not requre a cassette and has no movable parts. This is
extremely desirable for covert applications.
As with all analog recorders the speed invariably increases and decreases
slightly during recording. This increase and decrease is known as wow and
flutter. Wow and flutter are literally unmeasurable with DAT and
nonexistent with semi-conductor recorders.
The semi-conductore recorder has a signal to noise ratio (S/N) greater than
90db, which is so quiet that a noise reduction sytem is not requred. By
way of comparison, an analog tape recorder such as the Nagra has a S/N
ratio around 70db; perhaps in the low 80db range with noise reduction.
Please keep in mind that the S/N ration can be improved on any recorder,
using a tape, by running it at a faster speed. However, faster speeds
mean more frequent reel changes resulting in less recording time.
Digitals recorders have an extremely wide dynamic range, while this range
is much narrower for analog tape recorders.
DSP FOR NOISE CANCELLATION
The analog meethod of cancelling the unwanted noise is to pass the signal
through a filter that tends to supporess the noise while leaving the signal
relatively unchanged. Filters for this analog procecure are fixed. When
using fixed filters a prior knowledge of both the signal and noise is
requred. This method is extremely difficult to conduct during real time
surveillance operations, especially, when there is a variety of background
noises.
Digital adaptive filters have the ability to adjust their own parameters
automatically. This method requres little or no prior knowledge of he
signal or noies characteristics. The adaptive noise cancellation processor
uses a primary input containg the noisel The reference signal (noise) is
adaptively filtered and subtracted from the primary (subject and noise)
input to obtain a processed signal. This method can be demonstrated by
using a Frequency of Time Domain Adaptive Processor (FDAP or TDAP) in
a real time or recorder audio intercept. An example of a real time audio
intercept using a TDAP is depicted in the block diagram in Figure 3.
CONCLUSION
Digital signal processing holds a great potential for the technical
investigator of the 90's. DSP devices can be used to improve the quality
of audio intercepts both in the field and at the audio laboratory.
|
|
|
