Note: Descriptions are shown in the official language in which they were submitted.
2183868
SIREN DETECTOR
Field of the Invention
This application pertains to an improved siren
detector for detecting siren sounds which precess with
known characteristics within a selected frequency band. By
detecting siren sounds emitted by an emergency vehicle, the
siren detector facilitates preemptive control of traffic
lights to enable a vehicle equipped with the appropriate
siren to pass through an appropriately equipped intersec-
tion on a priority basis.
Background of the Invention
The prior art has evolved various ways of con
trolling or ~~pre-empting~~ vehicle traffic lights to stop
traffic at an intersection so that an emergency vehicle may
pass unimpeded through the intersection on a priority
basis. One technique involves the placement of a special
transmitter on each emergency vehicle which is to be
allowed priority passage through intersections. The
traffic light controllers at each preemptable intersection
are equipped with a receiver which receives signals trans-
mitted by the transmitter and there upon actuates the
traffic lights to stop the normal flow of traffic. However
this technique is relatively expensive and is cumbersome in
that the personnel in the emergency vehicle must manually
actuate the transmitter in order to control the traffic
light.
Traf f is light controllers at preemptable inter
sections have also been equipped with detectors capable of
detecting flashing lights (normally special strobe lights)
mounted on each emergency vehicle which is to be allowed
priority passage through the preemptable intersections. In
essence, this is similar to the system mentioned in the
preceding paragraph, in that the emergency vehicle light
replaces the special transmitter. The system does however
enjoy something of a cost and utility advantage over the
system mentioned in the previous paragraph, since emergency
vehicles are normally equipped with flashing lights which
2i 838b8
_ - 2 -
are actuated in emergency situations. However, the cost
advantage diminishes if special lights must be provided in
order to actuate the detector circuitry which interfaces
with the traffic signal controller. Moreover, the inven-
tors believe that such systems are susceptible to false
alarm triggering because, so far as the inventors are
aware, there are no regulations regarding the use of
flashing lights on non-emergency vehicles. Accordingly,
private vehicles may disrupt such systems by equipping
their vehicles with flashing lights for the express purpose
of actuating the detectors which interface with the traffic
light controllers. Perhaps a more serious situation is one
in which flashing lights used in advertising signs, commer-
cial window displays, and decorative lighting may falsely
trigger the detector. This is most prominent in dense
urban areas, which is precisely the area that the preemp-
tive traffic light signalling system is meant to provide
reliable triggering and afford the emergency vehicle the
shortest possible response time to its destination.
In the inventors' view a better solution is to
devise circuitry capable of detecting the sounds produced
by emergency vehicle sirens. There is clear cost advantage
to this approach, in that emergency vehicles are conven-
tionally equipped with sirens (ie. the emergency vehicles
do not need to be equipped with additional special purpose
equipment) and a utility advantage in that such sirens are
normally activated in emergency situations (i.e. no separ-
ate manual actuation of additional special purpose equip-
ment is required). A further advantage is that regulations
do exist which prohibit the use of sirens on non-emergency
vehicles.
The prior art has evolved a number of circuits
for detecting siren sounds (see for example published
European Patent Application No. 318,668 and United States
Patent No. 4,956,866). However, the inventors consider
these to be problematic in that they are susceptible to
false alarm triggering by sounds emanating from sources
2183868
- 3 -
other than emergency vehicle sirens. They also provide
unreliable detection of siren signals that have a relative-
ly long period as well as very long detection times. The
present invention provides an improved siren detector for
reliably detecting siren sounds within a selected frequency
band and having superior immunity to false alarm triggering
by sounds emanating from sources other than emergency
vehicle sirens, and having superior ability to detect siren
sounds in the presence of high ambient noise levels, and
detecting siren signals which have a relatively long period
in a short period of time.
The invention is based on the observation that
the majority of siren sounds are characteristic of a
frequency modulated (or FM) waveform in which the frequency
is modulated with a very characteristic and periodic
waveform. By using techniques common to radio receiver
engineering, it is possible to use traditional FM detection
schemes to obtain a very accurate estimate of the frequency
modulation waveform. This allows simple pattern recogni-
tion to be applied to this modulation waveform and accurate
recognition of various waveform patterns to be made. In
addition, the ability of the FM detection scheme yields a
great increase in the ability of this invention to detect
sirens in very high noise levels. With the low cost, high
degree of functional integration, and ease of reprogramming
for different algorithms and parameters, Digital Signal
Processing (DSP) techniques lend themselves to such a siren
detection system.
Summary of the Invention
The invention provides a siren detector for
detecting siren sounds which precess at known rates within
a selected frequency band to facilitate preemptable control
of traffic light signals to enable an emergency vehicle to
pass through a traffic intersection on a priority basis.
A transducer detects the siren sounds and produces a
corresponding electrical sound output signal. The sound
283868
- 4 -
output signal is then filtered to produce an antialiased
output signal which prevents aliasing in a subsequent
analog to digital conversion process. The antialiased
output signal is then bandpass filtered to reject signals
outside the selected frequency band. A limiter-discrimina-
tor then performs a constant amplitude scaling of the
amplitude components of the filtered antialiased output
signal to produce an output signal having frequency and
amplitude-scaled components and subsequently removes the
amplitude-scaled component, leaving only the desired fre-
quency component.
A click filter removes impulsive noise components
from the frequency component output by the limiter-
discriminator, to produce a filtered discriminator output
signal. A sound level detector responsive to the bandpass
filtered signal produces a sound level signal indicating
that a sound level within the selected frequency band at
the input transducer exceeds a selected sound intensity
level. A squelch detector is provided to indicate whether
a signal to noise level within the selected frequency band
at the input transducer exceeds a selected signal to noise
level.
A detector responsive to the frequency component
output by the limiter-discriminator measures the siren
sound's period and indicates whether the period is within
a selectable range. Another detector responds to the
frequency component output by the limiter-discriminator by
measuring the siren sound's frequency and indicating
whether the measured frequency is within a selectable
range. A means is also provided for measuring the siren
sound's rate of change of frequency and for indicating
whether the measured rate of change of frequency is within
a selectable range.
A means is also provided to determine a correla
tion coefficient providing a measure of correlation between
the precession rate of the siren sound and a straight line,
X183868
- 5 -
and for indicating whether the correlation coefficient
exceeds a selectable value.
A preempt control means produces a preempt output
signal for preempting control of the traffic light signals
when the siren sound increases in level above a selectable
threshold and for deactivating the preempt output signal
when the siren sound decreases in level below a selectable
threshold. The preempt output signal is held in an enabled
state for a selectable period of time.
The invention may be implemented as a program-
mable signal processor operated according to a computer
program. The programmable signal processor may have a com-
munications port allowing the computer program or user
selectable operating parameters to be externally loaded
from an external source, which may be remotely located.
Brief Description of the Drawings
Figure 1 is a block diagram illustrating the
basic operation of the siren detector according to the
invention.
Figure 2a and 2b are diagram illustrating the
basic configuration of a four channel siren detector at a
street intersection, and the configuration of a plurality
of siren detectors.
Figure 3 is a block diagram illustrating the
limiter discriminator of the siren detector according to
the invention.
Figure 4a, 4b, and 4c are diagrams illustrating
the ideal characteristic signals of three of the many
common types of siren sound which are detected when pro
cessed in accordance with the preferred embodiment of the
invention.
Figures 5, 6, and 7 are diagrams illustrating the
typical actual characteristics of three of the many common
types of siren sound which are detected when processed in
accordance with the preferred embodiment of the invention.
These are the yelp, high-low and wail respectively.
X183868
- 6 -
Figure 8 is a diagram illustrating the effect of
the click filter in removing the FM clicks from the re-
ceived signal when processed in accordance with the pre-
ferred embodiment of the invention.
Figure 9 is a diagram illustrating the operation
of the median filter, used as the click filter.
Figure 10 is a detailed diagram of a generalized
siren detector used for classifying a sound as being one of
a number of desired siren types.
Figure 11 is a block diagram of a noise operated
squelch detector.
Figure 12 is a diagram depicting the means for
measurement of the waveform period for yelp and high-low
sirens.
Figure 13 is a diagram depicting an alternate
means for measurement of the high-low siren.
Figure 14 is a diagram depicting the means by
which a wail siren sound is detected using the linear least
squares fit of a short line segment to the sampled siren
data.
Figure 15 is linear correlation coefficient plot
for a linear least squares fit to a wail siren. This is
the "linearity coefficient" output of the slope detector.
Figure 16 is the signal slope output of the slope
detector, which~gives the rate of change of frequency of
the siren signal, for a wail siren.
Figure 17 is a block diagram of the siren de-
tector showing the preferred embodiment.
Detailed Description of the Preferred Embodiment
Emergency vehicle sirens commonly emit sounds
which precess between two frequencies, the minimum and
maximum frequencies, with known repetition rates and
characteristics. Three of the more common siren sound are
commonly referred to as the yelp, high-low, and wail. The
ideal characteristics are shown in Figures 4a, 4b, and 4c
respectively. Ideally, the siren has a constant intensity
~":..s.
~ia3a6a
-
as the signal precesses according to these siren character-
istics, and others. A yelp siren sound typically has a
minimum frequency of 400 Hz, a maximum frequency of 1400
Hz, and a repetition rate of about 3 Hz. A high-low siren
sound typically has a minimum frequency of 400 Hz, a
maximum frequency of 600 Hz, and a repetition rate of about
1 Hz. A Wail siren sound typically has a minimum frequency
of 400 Hz, a maximum frequency of 1400 Hz, and a repetition
rate of about 0.25 Hz. Other siren sounds exist, and new
ones may be defined, which may also be detected by this
invention using the method described in this invention.
Fig. 1 is a block diagram which illustrates the
basic operation of a siren detector constructed in accord
ance with the invention. A brief overview of the invention
will first be provided with reference to Fig. 1. A de
tailed description of the preferred embodiment will then be
provided.
With reference to Fig. 1, the siren detector
utilizes an input transducer 1 to detect sound energy and
convert those to electrical signals suitable for processing
by the siren detector. These electrical signals are am-
plified to some nominal level for processing. The preamp-
lifier 2 is followed by an anti-aliasing filter 3 prior to
the analog to digital convertor 4 which converts these
analog electrical signals to a digital form for subsequent
processing. An analog to digital convertor with a resol-
ution of 12 to 16 bits and a sampling rate of 8.0 kHz has
been found to be suitable for processing the wail, yelp,
and high-low sirens described so far. A band pass filter
5 with a passband from about 300 Hz to 1800 Hz has been
found to suitable for wail, yelp, and high-low sirens. The
sampling rate would have to be increased above 8.0 kHz if
sirens with maximum frequencies much higher than those
discussed so far are to be sampled without aliasing. The
digital bandpass filter 5 is used to remove spectral energy
outside of the band found in the wail, yelp, and high-low
detectors. A passband of 300 Hz to 1800 Hz has been found
- 21~386~
-8_
to suitable for these sirens . Those skilled in the art
will realize that the bandpass filter 5 can be combined
with the phase splitter required for the limiter-discrim-
inator 6 described in Fig. 3, thus reducing the overall
complexity of these two functions. The limiter-discrim-
inator 6 measures the instantaneous frequency of the
received signal and the magnitude of that signal. Because
the spectral components of the frequency output of the
limiter-discriminator, representing the precession of the
siren signal, are so low for wail, yelp, and high-low
sirens, the output sample rate of the limiter-discriminator
vastly exceeds that required. For this reason, the limit-
er-discriminator output signal sampling rate is reduced by
the decimator 7 to a much lower sample rate. A decimation
of 8.0 kHz to 40 Hz has been found to be suitable. Since
the actual spectral content of the sirens variation of
frequency with time, as shown in Figs. 5, 6, and 7, is
typically less than about 15 Hz, the sample rate after the
low pass filter in the decimator need only really be
greater than about 30 Hz. This sample rate reduction
greatly reduces the processing demands of the subsequent
steps.
Another key advantage of this low pass filter
operation is that it allows the limiter-discriminator
detector to be operated essentially as a wideband frequency
modulation detector. This allows the great improvement in
siren detectability over conventional means. As is the
case with conventional FM receivers of the type discussed
by Jakes in "Microwave Mobile Communications" (John Wiley
& Sons, 1974 ISBN 0-471-43720-4) it can be shown as the
ratio of the input signal bandwidth at the input transducer
1 to the baseband output of the limiter discriminator 6
increases, the baseband output signal to noise ratio
increases for the same input signal to noise ratio. The
input bandwidth of the detector is defined by the input
signal bandpass filter, which is about 1500 Hertz, and the
low pass filter following the limiter-discriminator, which
2183868
- 9 -
is about 15 Hertz. The performance gains of wideband
versus narrowband FM detection are discussed in great deal
by Jakes (supra). It is this detection scheme which allows
sirens to be detected reliably in condition with signal to
noise ratios as low as -2 dB, whereas conventional detec-
tion means typically require a signal to noise ratio of
about 6 dB or higher. This invention provides approximate-
ly 8 dB gain over conventional means.
It is a characteristic of discriminator type
detectors that an FM modulated waveform, such as the siren
sounds, produce impulse noise or "clicks" when the signal
to noise ratio of the sound is low. This occurs when a
siren sound is some considerable distance from the input
transducer, or the background sound level in the vicinity
of the input transducer is very high. In any case, these
"clicks" create a problem when trying to classify siren
sounds belong to one class of a number of classes of
sirens. In Fig. 7, the actual limiter-discriminator fre-
quency output signal for a wail siren with a low signal to
noise ratio is shown. The clicks are clearly evident at
about 1.5 seconds and 6.3 seconds elapsed time in the
figure. A click filter 8 as shown in Fig. 1 can very
effectively remove these clicks from the limiter-discrim-
inator frequency output signal. The same input signal in
Fig. 7 when processed by this click filter results in a
median filter output as shown in Fig. 8, where the clicks
are seen to be removed. It has been found that a "Median
Filter" with a length of 9 samples or about 0.225 seconds
time duration is quite effective at removing these clicks.
Longer duration Median filters could be used, but they show
no substantial improvement in performance.
The output of the click filter 8 in Fig. 1 serves
as an input to a plurality of detectors. In this case,
they are yelp detector 9, High-Low detector 10, and wail
detector 11. One of more "Other Siren Detectors" 12 may be
added to detect additional siren types, or replace any or
all of the yelp, high-low, and wail siren detectors. These
X183868
- 10 -
detectors determine if the variation of the signal fre-
quency with time meets a number of criteria which classify
it as one of a number of siren types which the siren
detector has been configured to detect. The outputs) of
these detectors serve as one of a number of inputs to the
Preempt Detection Logic 15. The preempt detection logic
uses the outputs from the siren detectors 9, 10, 11, 12,
the squelch detector 13, and the sound level detector 14
to determine if the sound detected meets the siren detec-
tion criteria. If they do meet the selection criteria,
then the PREEMPT signal to the traffic light controller is
enabled.
The output of the Bandpass Filter 5 in Fig. 1,
typically with a passband from about 300 Hz to about 1500
Hz., is a signal whose amplitude is a function of the siren
loudness or level at the input transducer 1. Since sirens
maintain an approximately constant output level and the
sound level at 1 increases with decreasing distance between
the siren and the input transducer, the signal level at 5
is a function of the distance between the input transducer
and the siren. The signal at 5 is input to the Sound Level
Detector 14 which measures the magnitude of the that signal
and compares it against a preset level threshold. If the
magnitude of the signal at 5 exceeds the level threshold,
it enables the output of the Sound Level Detector. If the
magnitude of the signal at 5 does not exceeds the level
threshold, it disables the output of the Sound Level
Detector. The output of the sound level detector serves as
one of the inputs to the Preempt Detection Logic 15.
In some situations the ambient sound level from
sources other than sirens, such as that due to traffic
noise from tires, engine noise, industrial noise, aircraft
engine noise, etc., may be so loud that these levels exceed
the detection level threshold of the Sound Level Detector
14. In this situation, the output of the Sound Level
Detector 14 would always be enabled and the siren would
cause the Preempt Detection Logic 15 to cause a PREEMPT
s
X183868
- 11 -
signal sooner than is desired. By utilizing a conventional
squelch detector, an additional signal which is a function
of the signal to noise ratio is available. The squelch
detector is configured such that a threshold signal to
noise ratio must be exceeded before the squelch detector
output is enabled to indicate this detection criteria has
been met.
The PREEMPT detection logic 15 uses combinations
of the squelch detector 13 output in addition to the siren
detector functions, shown in 9, 10, 11, and 12 and the
sound level detector 14 of Fig. 1. In normal urban and
suburban situations, the PREEMPT detection logic 15 would
only enable the PREEMPT output to the traffic light con-
troller when; (a) the sound reaching then input transducer
1 meets one of the valid siren selection criteria of siren
detector functions shown in 9, 10, 11, and 12, and (b) the
sound reaching then input transducer 1 exceeds the detec-
tion threshold criteria of the sound level threshold
detector 14. For very noisy environments, the PREEMPT
detection logic 15 would only enable the PREEMPT output to
the traffic light controller when; (a) the sound reaching
then input transducer 1 meets one of the valid siren
selection criteria of siren detector functions shown in 9,
10, 11, and 12, and (b) the sound reaching then input
transducer 1 exceeds the detection threshold criteria of
the sound level threshold detector 14, and (c) the signal
to noise ratio measured at the output of the limiter-dis-
criminator 6 measured by the squelch detector 13 exceeds a
squelch detection threshold.
Fig 2 (a) shows a typical installation with a
traffic light 26, four input transducers 21, 22, 23, and 24
mounted such that they are optimized for detection of sound
from one of the four streets which approach the traffic
signal 26. The output signals from these transducers go to
a four channel siren detector 20 which processes the
signals from the input transducers. If an emergency
vehicle 25 approaches in the direction of input transducer
2183868
- 12 -
24, the channel in the siren detector processing that
signal will indicate a PREEMPT signal to the traffic Light
Controller 30 for that direction of the traffic light 26
using the traffic light preempt line 31, and/or the pedes-
trian control preempt line 32. The Traffic Light Control-
ler could then be configured to give the emergency vehicle
25 priority access to the intersection. As indicated in
Fig. 2(b), the siren detector can consist of a plurality of
siren detector channels ranging from 1 to many. However,
4 channels is the most common. Single channel detectors
could be to control lights at the driveway to fire halls,
police compounds, pedestrian controlled lights. etc.
Fig. 3 shows one means for realizing a lim
iter-discriminator. The input signal is split into its
real and imaginary components by the phase sputter 40.
The complex conjugate and first derivative of the phase
splitter output are formed by 41 and 42 respectively. The
product of the complex conjugate and first derivative is
taken, as well as multiplied by -j - -~-1. The real part
of this product is taken by 44. The power of the input
signal is determined by taking the magnitude of the phase
splitter output in 46, and then squaring this signal in 47.
The frequency of the input signal is then calculated by
dividing in block 45 the output of 44 by the output of 47.
The output of 47 also serves as the input to the sound
level detector 14 in Fig. 1.
Fig. 4 (a), (b), and (c) show the ideal frequency
versus time characteristics of the three most common
sirens, these being the yelp siren, high-low siren, and
wail siren respectively. In actual practice, the sirens
characteristics are quite different. Fig. 5 shows the
frequency versus time characteristic of a yelp siren. Fig.
6 shows the frequency versus time characteristic of a
high-low siren. Fig. 7 shows the frequency versus time
characteristic of a wail siren. In these three examples,
the frequency was measured with actual sirens using the
limiter-discriminator shown in Fig. 3.
2j83868
- 13 -
The Median filter is commonly used in image
processing to remove impulsive noise. It operates by
assembling an odd number of sequential data samples,
sorting the samples in ascending or descending order, and
then extracting the medial value. It operates in much the
same way as sliding window finite impulse response filter,
except that it is quite non-linear in nature. The use of
the click filter is necessary for the detection of siren
sounds where the signal to noise ratio is low. Fig 8 shows
the effect of the median filter on an actual wail siren
signal having a low signal to noise ratio. The input
signal is shown in Fig 7. Using the example of the median
filter shown in Fig. 9, the operation of the median filter
can be easily demonstrated. The input samples 50 are
serially shifted into the input shift register 51. They
are sorted in ascending ( or descending ) order by the
sorter 52 and reassembled in ascending ( or descending )
into the output register 53. From the output register 53,
the medial value is taken and used as the output. In the
example shown, the sampled data sequence in the register 51
is 1, 4, 6, 2, 9, 8, 5, 7, and 3. From this sequence, the
median filter selects 5 as the medial value. If a new
input sample with a value 11 was input into the shift
register 51, the end value 3 would be discarded and the
input shift register 51 contents would become 11, 1, 4, 6,
2, 9, 8, 5, and 7. These would result in the output shift
register contents becoming 1, 2, 4, 5, 6, 7, 8, 9, il after
sorting. The medial value output by the filter 54 would be
6 in this case.
Three basic types of sirens detectors are used
for the detection of most sirens . The main obj ective of
these schemes is to provide a low probability of false
detection, fairly fast detection and classification time of
about 2 to 3 seconds maximum, and sufficient flexibility to
accommodate variations in the siren characteristics. A
common core siren detector is shown in Fig. 10, serving as
2183868
- 14 -
the basis for the detection of yelp, wail, high-low, and
other siren types.
The first of these is the most general and is
suitable for yelp siren, although other siren types could
also be detected. It simply sets a frequency threshold
comparator 61 with a frequency threshold f~.esh midway between
the minimum and maximum frequencies expected for a yelp
siren, which is about 900 to 1000 Hertz. The period
between times when the increasing frequency wave shape
crosses the threshold for two successive threshold crossing
is measured by 62. If this period falls within the user
selected range for valid yelp sirens which is typically
0.27 seconds to 0.40 seconds, and the frequency of the
siren signal is greater than a selectable minimum frequency
f";", and less than a selectable maximum frequency f~x, a
counter is incremented. The frequency comparators 63 and
64 are used for the purpose of frequency comparison. If
the next period is measured to be within the user selected
region, the counter is incremented again. If the next
period is measured to be outside of the user selected
range, the counter is decremented. The counter minimum
value is 0. If the counter level exceeds a user selected
threshold, typically 3 or 4 for reliable detection, then
the yelp detector output is enabled to indicate that a
siren meeting the yelp detection has been detected. It
should be apparent that the sense of the change in fre-
quency from an increasing in time sense to a decreasing in
time sense in relation to the frequency threshold crossings
is also possible within the context of this invention.
This means may also be used for the high-low siren type,
since this siren type is characterized by its periodic two
frequency characteristic. The period measurement technique
is shown in Fig. 12.
The second of these is also suitable for high-low
siren, although other siren types could also be detected.
It simply sets a frequency difference threshold midway
between the difference of the minimum and maximum fre
2183868
- 15 -
quencies expected for a high-low siren, which is about 100
to 150 Hertz. The frequency comparator 61 is then used to
determine if the step in frequency between the low tone and
the high tone exceeds some threshold f~,iesh. The period
between times when the increasing frequency wave shape
crosses the threshold for two successive increasing fre-
quency crossings is measured. If this period falls within
the user selected range for valid yelp sirens which is
typically 1.00 seconds to 1.3 seconds, and the frequency of
the siren signal is greater than a selectable minimum
frequency f,n;" and less than a selectable maximum frequency
fm~, a counter is incremented. The frequency comparators
63 and 64 are used for the purpose of frequency comparison.
If the next period is measured to be within the user
selected region, the counter is incremented again. If the
next period is measured to be outside of the user selected
range, the counter is decremented. The counter minimum
value is 0 and typically has a maximum value of less than
20. If the counter level exceeds a user selected thresh-
old, typically 3 or 4 to reliable detection, then the
high-low detector output is enabled to indicate that a
siren meeting the high-low detection has been detected. It
should be apparent that the sense of the change in fre-
quency from an increasing in time sense to a decreasing in
time sense in relation to the frequency threshold crossings
is also possible within the context of this invention. The
period measurement technique is shown in Fig. 13.
The third siren detector type is for the wail
siren. This siren type is characterized by a very long
period of between 4.8 and 7.2 seconds. It is readily
apparent that if three to four complete cycles of a wail
waveform were to be detected before the wail detect output
were enabled, a detection time of about 15 or 20 seconds to
22 to 29 seconds would be required. This greatly exceeds
the desired 2 to 3 seconds detection time. In fact, a
siren equipped vehicle could easily be passed the intersec-
tion before the siren would have been detected. This
~1838b8
- 16 -
highly undesirable situation is alleviated by observing the
fact that the frequency characteristic is more or less a
triangle wave with fairly straight portions to the curve.
The Wail siren detector uses this fact, and uses a short
duration sliding window of about 1.0 seconds in duration to
perform a linear least squares fit to the sampled frequency
data. A linear equation of the form:
f = mt + b
is fit to a 1.0 second sequence of data samples, number 40
for the siren detector being discussed. In this equation,
f is the frequency, t is the time, m is the slope of the
line or rate of change of frequency, and b is the intercept
frequency at t - 0.0 seconds. Also calculated is the
linear correlation coefficient of the fit between the
straight line segment and the samples of data. One way of
calculating this linear correlation coefficient for N
samples of data, with N being 40 in this case, is using the
following equation:
N~t; fi - Frtq ~. f;
r =
[ NF.ti2 - (F~t;) 2 ] 1~ [ NF.f;2 - (Efi) 2 ]''~
where f; is the frequency taken at time ti and N is the
number of samples used in the linear fit. The value of r
ranges from 0 where there is no correlation, to ~1 where
there is complete correlation. The sign of r in this case
is the same as that of the slope m, but it is only the
magnitude r that is important and not the sign.
This linear least squares fit to the waveform and
the frequency at any part of the waveform provide three
classification criteria for the wail siren. These criteria
are; (1) the frequency of the waveform must be with the
user specified minimum and maximum frequencies as deter-
mined by comparators 63 and 64, (2) the rate of change of
the frequency with time or slope of the straight line
portion of the curves must fall within two user defined
ranges, typically between ~300 Hz/sec to ~ 500 Hz/sec, as
determined by the slope detector 65, and (3) the goodness
2183868
- 17 -
of fit or correlation coefficient of the piecewise linear
line segment to the frequency waveform as determined by the
slope detector 65, with the magnitude of a good linear
correlation coefficient typically being between 0.95 and
1Ø If the siren meets all three of these criteria, it
can be reliably classified as a wail siren type. Typical
detection times using this technique are the order of 2 to
3 seconds, making it as reliable as the yelp siren detec-
tion technique. The slope measurement technique is shown
in Fig. 14. The slope m of the wail siren sound shown in
Fig 8 is shown in Fig. 15, and the linear correlation
coefficient r is shown in Fig. 16. In this example, the
sample rate was 40 Hertz and 40 sample points were used for
the linear fit. This fit was performed at a rate of 40
Hertz.
One common type of squelch detector is based on
a noise operated squelch detector. This detector provides
a signal which is a function of the baseband SNR of the
limiter-discriminator output. It is described in detail by
Rhode and Ulrich in "Communications Receivers: Principles
& Design", McGraw-Hill Book Company, 1988. The operation
of these noise detectors is based on the fact that as the
carrier to noise ratio increases, the baseband noise energy
density decreases. This detector used for this purpose is
shown schematically in Fig. 11. The output of the 1.5 kHz
to 1.8 kHz bandpass filter is "full-wave rectified" by the
Absolute value block. This output is then filtered by a
simple low pass filter with a bandwidth of about 10 Hertz.
The output of this filter is then decimated to a rate of 40
Hertz, reducing the subsequent processing rates. The
decimated output, which is a function of the signal to
noise ratio of the squelch input signal, is then compared
against a user selected threshold and the threshold de
tector output enabled when the input signal is below the
threshold level.
Those skilled in the art will recognize that the
siren detector described in this invention is ideally
~ ~ a3a6a
- 18 -
suited for implementation in a programmable computing
device or digital signal processor. This has the many
advantages over analog implementations, such as little if
any effect of temperature on the performance, ease of
adapting the siren detector to new siren sounds by repro-
gramming rather than modifications to the hardware, the
ability to~remotely reprogram the siren detector for new
siren sounds, the ability to remotely control the siren
detector, etc. This preferred implementation is shown in
Fig 17. The input signals from the input transducers are
input to the Analog Input Signal Protection, Amplification,
and Filtering section 80 to provide electrical transient
protection and signal conditioning. The signal processor
81 performs the analog to digital conversions and all of
the processing functions described in this invention.
Status indicators provide feedback to users as to the
performance of the siren detector, detection of valid siren
sounds, siren type, channel number activated, etc. Parame-
ter input selectors 84 are provided to allow adjustment of
the siren detection parameters locally. An External
Programming and Control Input Port 85 is provided to allow
local or remote reprogramming of the siren detector to
update the software control program, or to locally or
remotely change the siren detection parameters.
As will be apparent to those skilled in the art
in the light of the foregoing d~i.sclosure, many alterations
and modifications are possible in the practice of this
invention without departing from the spirit or scope
thereof. Accordingly, the scope of the invention is to be
construed in accordance with the substance defined by the
following claims.
.4
