Method and apparatus for indicating activation of a smoke detector alarm

This application claims priority from US Provisional Application Serial No. 60 / 415,127 filed on October 2, 2002. The full text of this provisional application is incorporated herein by reference.

<u style="single">Background of the invention</u><u style="single">Field of invention</u> The present invention relates to methods and devices for sensing the operation of an audible smoke detection alarm.

<u style="single">Technical background</u> In the event of a fire, people in the building may only take a few minutes to evacuate without damage. Due to the potentially short evacuation time, it is essential to give sufficient warning to those in the burning building. Most devices sold by disaster prevention companies rely on audible alerts to alert people in residential buildings. Unfortunately, these devices do not help the hearing impaired. Therefore, there is a need for a device that provides sufficient protection for the hearing impaired in the event of a fire emergency.

<p> Known for this technology is a device that uses visible signals to warn hearing-impaired people in the event of a fire. Examples of such devices are described in US Pat. Nos. 4,227,191 and 4,287,509. These devices combine a sensor and a visible alarm into a single device. Other visible warning devices are disclosed in US Pat. No. 5,012,223. The device senses sound from a remote smoke detector and emits light in response. These visible alarm devices have the serious drawback of not being effective in alerting the hearing impaired during sleep.</p><p> To meet this need, systems that combine tactile stimuli (eg, vibrators and bed shakers) have been proposed. One such device is described in US Pat. No. 4,380,759. This device includes a vibration sensor installed adjacent to the smoke detector. When the smoke detector is activated, the vibration from the audible alarm activates the vibration reed, causing a slight irritation to the skin. Such devices require the transmission unit to be installed so that the user often comes into physical contact with smoke detectors on the ceiling or other inaccessible areas (especially when the device is temporary, such as in a hotel room). (When used only for), it is troublesome to use. Other devices for the hearing impaired (eg, the device disclosed in US Pat. No. 5,917,420) involve signaling from a sensor to a furniture shaker or to another tactile stimulator. While such equipment is a step in the right direction, it is usually quite expensive and, more importantly, requires special hardware.</p><p> U.S. Pat. No. 5,651,070 describes a warning device that "listens" to the sounds produced by devices such as doorbells and smoke detectors and activates a wristwatch-shaped tactile stimulator. The device records the desired audible alarm and constantly compares the ambient noise picked up by the microphone with the recorded alarm. A 4-bit comparator is used to declare the match, but it is not disclosed what criteria are input to the comparator to declare the match. In addition, this device is difficult to use in that the user needs to record the desired sound before use. This can be a problem, for example, if a user enters a hotel room late at night. This is because activating the smoke detection alarm for the purpose of recording is a nuisance to other guests.</p><p> In addition to the above concerns, the issue of false alarms is an important issue in such devices. Sensing the activation of a smoke detector's audible alarm is a relatively simple matter, but being able to detect such an alarm without accidentally detecting other devices commonly found in the home is completely different. It's a problem. Such devices can produce persistent wideband noise (eg vacuum cleaners and cooking mixers), and intermittent sounds with distinct frequencies that are at or near the frequency of smoke detectors and alarms. May also be produced (for example, alarm clocks, phones, cell phones, etc.). It is well known that users quickly learn to ignore alarm devices that cause a large amount of false alarms and make such devices useless.</p>

<p><u style="single">Outline of the invention</u> The above-mentioned problems are solved to a considerable extent by the present invention providing a method and a device for detecting the presence of a sound corresponding to both a continuous smoke warning sound and an intermittent smoke warning sound. Determining an alarm involves an algorithmic analysis of at least two parameters of the sound data and triggering the alarm device after sensing the signal of a smoke detection alarm. In a preferred embodiment, an alarm device is used to provide a tactile stimulus.</p><p> In a preferred embodiment, a series of temporally spaced samples of the temporally repeating sound pattern of the smoke detector's audible alarm sound are obtained over a sufficiently long period of time to include at least one entire period. The two parameters analyzed for each sample are the frequency and amplitude of the maximum sound of that sample. Both of these parameters must match the desired temporal pattern for the audible alarm to be declared. In an embodiment of the invention directed at the sensing of intermittent audible smoke sensing alarm sounds, the pattern fluctuates over time. Persistent audible smoke detection In the embodiments directed to the detection of alarm sounds, the pattern does not fluctuate over time.</p><p> In other aspects of the invention, which may include, but may not include, a warning device for the hearing impaired, the initiation of the algorithmic analysis is a lower power relative to the amount of power consumed by the algorithmic analysis. Triggered by a detection algorithm designed to take advantage of the quantity. This is especially important for battery-powered devices.</p><p> In yet another aspect of the invention, the above-mentioned detector is integrated into a conventional smoke detector to provide a smoke detector that senses both smoke and alarms from other smoke detectors. In an alternative embodiment, the smoke detector includes a transmitter / receiver, which sends an activation signal to a nearby smoke detector when smoke indicating a fire is detected, and also a nearby smoke detector. Receives an operation signal from. Upon receiving an activation signal from a nearby smoke detector, the smoke detector activates a warning device (audible, tactile, and / or visibility). Such embodiments are particularly useful when station-to-station (station-to-station) operation (eg, by building code) is required. This is because this embodiment provides a means to achieve station-to-station operation without the need for wiring.</p><p> Many of the present invention and its accompanying features and advantages are better understood in the context of the accompanying drawings with reference to the detailed description below, thus facilitating a more complete understanding of them. Will be obtained.</p>

<u style="single">Detailed explanation</u>The present invention will be described with reference to preferred embodiments of devices for sensing audible smoke sensing alarm sounds. Specific details are given to provide a complete understanding of the invention. The preferred embodiments described herein should not be understood to limit the invention. In addition, it makes it easier to understand Therefore, certain method steps are distinguished as individual steps, but these steps must not be considered to be distinct in their execution, nor to be order-dependent.

Prior to 1996, smoke detector audible alarms had different patterns depending on the manufacturer. Some audible alarms were persistent and others showed different intermittent temporal patterns. However, smoke detectors manufactured after 1996 are standardized by the National Fire Protection Association, which stipulates that smoke detectors emit audible alarm signals with the temporal pattern shown in Figure 1. You are required to comply with an NFPA 72. This pattern consists of three sets of short on / off periods followed by a longer off period. The length of the short on and off periods is specified as "on" for 0.5 seconds +/- 10%, followed by "off" for 0.5 seconds +/- 10%. Long off periods are specified at 1.5 seconds +/- 10%. According to tests conducted by the applicant with several smoke detectors, most, but not all, detectors meet this specification.

In addition to the temporal pattern in Figure 1, the NFPA72 further specifies that an audible alarm signal intended to operate in private mode must have: -At least 45dBA acoustic level at 10ft from audible equipment, or over 120dBA acoustic level at minimum listening distance, -A sound level that is at least 15 dBA higher than the average ambient sound level, or 5 dBA higher than the maximum sound level that has a length of at least 60 seconds, measured at 5 ft above the floor, whichever is greater. The NFPA72 further specifies that in sleeping areas the audible alarm signal must have: -At least 15dBA higher than the average ambient acoustic level, or 5dBA higher than the maximum acoustic level with a length of at least 60 seconds, or at least 70dBA acoustics, measured at the pillow level in the sleeping area. The higher of the levels.

NFPA72 does not specify the frequency of the audible alarm sound. However, the applicant has tested approximately 18 currently available smoke detectors and has manufactured since 1996 (including several manufactured before 1996) and is currently available on the market for smoke detectors. It was found that most of the instruments had a frequency peak whose audible alarm sound was clearly distinguished at about 3200Hz +/- about 10%, and also had a peak at the corresponding harmonic frequency. The frequency of 3200Hz corresponds to sounds that are easily audible to most people with normal hearing. Figure 2a shows the amplitude of the audible alarm obtained in a foam-lined box in a typical smoke detector, in this case the Fire Sentry 0914 model smoke detector. An exemplary plot of frequency is shown. Figure 2b shows a plot of the amplitude vs. frequency of the same sensor in an unlined box. Table 2 summarizes the results of the 18 smoke detectors tested by the applicant.

<tables num="1"><img file="JP4279255B2_D0001.tif" /></tables>

The number of pre-1996 smoke detectors still in operation is declining rapidly, as many smoke detector manufacturers recommend replacing smoke detectors in almost a decade. Therefore, the present invention is mainly described in the context of recognizing audible alarms for smoke detectors made in 1996 and newer. However, the present invention should not be understood to be so limited.

FIG. 3 shows a device 300 according to an embodiment of the present invention for sensing an audible alarm sound of a smoke detector and activating a second alarm device. The device 300 includes a microphone 310 for sensing ambient sounds. The microphone 310 is optionally connected to the amplifier 320 (shown by the dotted line in FIG. 3). The amplifier 320 is connected to an analog-to-digital (A / D) converter 330 for converting an analog signal from the microphone 310 into a digital signal representing ambient sound. The digitized sound data from the A / D converter 330 is input to the processor 340 for detecting an audible smoke detection alarm sound. Processor 340 may be a microprocessor, digital signal processor, or any other type of processor. Although processor 340 is shown in a single box in Figure 3, it should be understood that processor 340 can include multiple devices. In one embodiment, the processor 340 includes a dedicated device for computing the Fourier transform and a general purpose microprocessor. Memory 350 is connected to processor 340.

Processor 340 is powered by power 360. In a preferred embodiment, the power source 360 is a battery. Alternatively, the power supply may include a converter and rectifier for connecting to an AC power supply. In yet another embodiment, the power supply 360 powers the processor 340 from an AC power source if an AC power source is available, or from a battery if an AC power source is not available, in a manner well known in the art. May be adapted as such.

When the processor detects an audible alarm sound from the smoke detector, the processor 340 outputs a signal to the alarm device 370. The alarm device 370 is a tactile alarm in a preferred embodiment. Tactile alarms that can be used in the present invention include vibrating watches, vibrating pocket bells, and the Super Bed Vibrator sold under the SONIC ALERTS® trademark, with variable speed. There are bed shakers such as super bed vibrators. The super vibrator has a diameter of 3.5 inches and a thickness of 1.25 inches and is unbalanced. Has a large mass and a motor. This device can be placed under a mattress or pillow. Other tactile alarms can also be used.

FIG. 4 is a flow chart 400 of processing executed by the processor 340 in one embodiment of the present invention. In step 402, the microphone 310 is used to sample the ambient sound for a period of time. This time period is selected as 2 seconds in some embodiments. This period is chosen to exceed the long off period of 1.5 seconds specified in the NFPA72 standard. In step 403, processor 340 performs a Fourier transform on the digitized sound data output by the A / D converter 330. Then, in step 404, the data is filtered by erasing the data at frequencies below the low frequency threshold. Those skilled in the art will recognize that this filtering step can be replaced by inserting a high-pass analog filter prior to the A / D converter 330.

After the data has been filtered in step 404, in step 406 the maximum amplitude of the filtered data is determined. In step 408, if this maximum amplitude corresponds to the expected frequency of the smoke detection alarm, i.e. 3200Hz +/- 10% (2880 to 3520Hz), then step 410 enters the acquisition routine. After the acquisition routine is complete, or when the peak identified in step 406 is not at the correct frequency in step 408 (ie, when some device other than the smoke detector is making the loudest noise at that time), processor 340. Is delayed for a period of time in step 412 and step 402 is repeated. In a preferred embodiment, the delay period for step 412 is 10 seconds.

Flow of FIG. 4 The routine shown in FIG. 400 serves as a periodic monitoring routine to determine if a smoke detection alarm may be sounding. In this context, "periodic" means from time to time and includes executing routines at both fixed and variable intervals. The choice of a 2-second monitoring period and a 10-second delay period corresponds to a duty cycle of approximately 17%. As an alternative, shorter samples that are continuously spaced over a period longer than the 1.5 second "off" period (or the longest off period of the target temporal pattern) rather than a continuous 2 second monitoring period. May be used. Rather than continuous monitoring, periodic monitoring routines are performed to save power, which is especially important for battery-powered equipment. In another embodiment of the invention, such as a conventional household AC powered device from a wall outlet, the monitoring routine 400 may be omitted and the acquisition routine 410 may be executed continuously.

Details of monitoring routine 410 are shown in FIGS. 5a and 5b. This routine begins in step 502 by sampling the ambient sound for a short period of time. In some examples, this sampling period is 50 ms. The minimum length of the sampling period must be chosen so that the number of samples obtained is sufficient to obtain good frequency resolution from the Fourier transform of the data. Therefore, the minimum length of the sampling period depends on the hardware sample rate. In step 504, processor 340 transforms the sample into the frequency domain by performing a Fourier transform, and in step 506 the sample is highpass by excluding data below a certain frequency threshold (500 Hz in the preferred embodiment). Be filtered. Next, in step 508, the frequency of the peak amplitude is determined. If the peak amplitude is at the expected frequency of the smoke sensing alarm (3200Hz +/- 10%, ie 2880Hz to 3520Hz, as described above), in step 512 the corresponding bits are set in the peak frequency array of memory 350. Otherwise, the corresponding bit remains "0". Peak circumference The wavenumber array is preferably a one-dimensional array of bits, where the total number of bits is equal to the total number of samples (eg, 12) during the total sample period. After step 512 (or after step 510 if the frequency of the peak amplitude is not the expected frequency of the smoke detection alarm sound in step 510), the peak amplitude of the sample is recorded in memory at step 514.

As further described below, the peak frequency array is correlated with the peak amplitude array and then compared to the expected temporal pattern. This ensures that the peak frequency of each "on" period of the NFPA72 pattern matches the expected frequency of the smoke detector's audible alarm. It does not matter if the peak amplitude frequency of the "off" period is at the target frequency, and the "off" period with the peak frequency at the target frequency of the smoke detection alarm sound results in an undeclared match. is not it. This is because ambient noise may have a peak at the same frequency as the smoke detection alarm during the "off" period.

If the desired number of samples is not obtained in step 514, the processor 340 is delayed for a period of time in step 518 and another sample is obtained in step 502. In a preferred embodiment, the delay period is selected as 0.5 seconds from the start of the last sample period, which corresponds to a 0.5 second "on" period and a 0.5 second "off" period of the temporal pattern defined by NFPA72. In a preferred embodiment, the total number of samples was selected to be 12, which corresponds to a total sample period of 5.5 seconds, with an interval of 0.5 seconds between the sample periods. As shown in Figure 6, 5.5 seconds is because the three "on" and long "off" periods specified in NFPA72, as well as one additional "on" period, are reliably present within the total sample period. Corresponds to the minimum amount of time required for. FIG. 7 shows 12 sample periods superimposed on one possible audible alarm signal.

With reference to FIG. 5a again, once the desired number of samples has been obtained for the total sample period in step 516, the maximum amplitude of all samples during the total period is determined in step 518. Next, in step 520, an amplitude threshold is set based on the maximum amplitude. The amplitude threshold is used to determine the threshold amplitude corresponding to the "on" condition. In some embodiments, the amplitude threshold is selected to be 80% of the maximum amplitude. The amplitude threshold depends on the expected amplitude of fluctuations in the sound source (ie, the audible alarm sound of the smoke detector). Therefore, the amplitude threshold can vary depending on the range of "target" smoke detectors and the desired false alarm rate.

It should be noted that setting the amplitude threshold as a function of the maximum threshold is considered to be an important aspect of the false alarm control of the present invention. In this threshold selection, the maximum amplitude of the smoke detection alarm will vary as a function of the square of the distance between the device 300 and the smoke detector, so the maximum amplitude may not be known, but the smoke detection alarm is. during the "on" period to a constant level, the distance between the device 300 and the smoke detector because normally does not change, the maximum amplitude during the "on" period, through the normal would be the same, the recognition To do. It should also be noted that the amplitude threshold is chosen as a function of maximum amplitude, not as a function of maximum amplitude for noise, or for noise signals. This is because the background noise may change, so that the signal-to-noise ratio is not constant even if the amplitude of the smoke detection alarm is constant. This amplitude threshold plays an important role in preventing false alarms because it helps distinguish smoke detection alarms from other devices such as televisions that emit intermittent amplitude-varying sounds.

After the threshold amplitude is determined in step 520, in step 522, the corresponding bits are set in the peak amplitude array for each sample whose maximum amplitude exceeds the threshold amplitude. The peak amplitude array, like the peak frequency array, is preferably a one-dimensional array of bits, where the total number of bits is equal to the total number of samples (eg, 12) over the total sample period. As an example, assume that the maximum amplitude perceived during the total sample period is:

<tables num="2"><img file="JP4279255B2_D0002.tif" /></tables>

In this case, the maximum amplitude is 9.5. The amplitude threshold is 9.5 x 0.80 = 7.6. Therefore, only the amplitudes in samples 1,3,5,9 and 11 exceed the threshold, resulting in a peak amplitude array of 101010001010.

In some embodiments, there is a bottom value where the amplitude threshold cannot be set below that. This bottom value corresponds to a level 15 dB higher than the ambient noise level specified by NFPA72. However, this minimum value of 15 dB assumes that the correct number of smoke detectors are installed (ie, the smoke detectors are within the maximum permissible distance) depending on the building. In other embodiments, the minimum value is only 10 dB for such non-standard installations. This minimum amplitude threshold ensures that ambient noise does not trigger false alarms.

Then in step 524, the peak amplitude array is correlated with the peak frequency array. As mentioned above, the peak frequency array is an array of bits, with one 1 for each sample corresponding to a frequency with a maximum amplitude of 3200 Hz. In a preferred embodiment, this correlation involves bitwise ANDing the peak frequency array and the peak amplitude array. The correlated array is also a one-dimensional array with as many bits as the number of samples, at a frequency where the peak amplitude exceeds the amplitude threshold and AND (and) matches the expected frequency of the smoke detector, ie about 3200 Hz. One bit is set for each sample with peak amplitude. In a low noise environment, the peak frequency array and the peak amplitude array can match. However, that does not always happen. For example, tests conducted by the applicant have revealed that in a peak frequency array, bits can be set for samples that are not set in a peak amplitude array. This is believed to be due to the slight reverberation from the "on" period, which naturally has the same frequency as the smoke sensing alarm, sensed during the "off" period in the NFPA72 temporal pattern. Correlating the peak frequency array with the peak amplitude array prevents these echoes from being mistaken for "on" signals.

Then in step 526, the correlated array is compared to the temporal pattern of NFPA72. As mentioned above, the total duration of the sample is chosen so that there are at least four "on" durations of the NFPA72 temporal pattern. Three of the four "on" periods are separated by a 0.5 second "off" period, and the other "on" periods are separated by a 1.5 second "off" period. However, it is unclear whether the three "on" periods precede the fourth "on" period. Therefore, there are two possible powers corresponding to these four "on" periods. The turn looks like this (where 1 is "on" and 0 is "off"): (a) 100010101 or (b) 101010001. Therefore, each of these two possible 9-bit patterns is compared with the 12-bit correlated array starting at the first, second, third, and fourth bits of the 12-bit array (in other words,). The pattern can start with any of the first 4 bits of the correlation array). If any pattern is found in the correlation array in step 526, the warning device is activated in step 528.

It is important that, unlike the peak frequency array, the peak amplitude array shows "off" in the sample corresponding to the "off" period of the NFPA72 temporal pattern. ANDing the peak frequency array (where the state of the sample corresponding to the "off" period is not important) with the peak amplitude array makes it easy to compare the two arrays with the NFPA temporal pattern.

It should be noted that these two patterns contain more than one complete "duration" of the patterns specified in NFPA72. This means that a complete period of the NFPA72 pattern contains only three "on" periods. The fourth "on" period in the above pattern corresponds to the second period in the NFPA72 pattern, which is included to provide additional measures to prevent false alarms. In some embodiments, the fourth "on" period is not included in the target pattern.

If no pattern is found in the correlation array at step 526, then at step 530 the total number of trials is compared to the total number of correlation trial thresholds. In a preferred embodiment, the total number of correlation trial thresholds is selected to be 10. Therefore, processor 340 attempts to find a temporal pattern over a period of 5.5 seconds 10 times for a total of 55 seconds, or about 1 minute. If the total number of correlation trial thresholds has not been achieved in step 530, step 502 and subsequent steps are repeated. When the total number of correlation trial thresholds is reached in step 530, the subroutine ends and control returns to step 412 (FIG. 4).

As mentioned above, the device 300 can also be used to detect smoke detectors prior to 1996. For example, some smoke detectors prior to 1996 emit audible alarms in a temporal pattern consisting of a series of on and off pulses. To detect such smoke detectors, the sample rate is adjusted to correspond to the timing of the on / off pulses, and the pattern in step 526 is changed to a sequence of alternating 1s and 0s. Other pre-1996 smoke detectors emit a continuous sound. For these perceptrons, the pattern is changed to a sequence of ones. A sequence of 1s is used instead of a single 1 (ie, a single sample with an amplitude higher than the amplitude threshold), ensuring that the duration of each signal is short, short-term sound (eg television sound). ) Is long enough to prevent it from being mistaken for a smoke detector audible alarm.

FIG. 8 shows a smoke detector 800 according to a second embodiment of the present invention. The smoke detector 800 is suitable for wall or ceiling installation in a manner similar to conventional smoke detectors. However, as described in more detail below, the smoke detector 800 is adapted for remote tactile alarm activation and station-to-station activation.

The smoke detector 800 includes a microphone 810 connected to the amplifier 820. The amplifier 820 is connected to the A / D converter 830. The A / D converter 830 is connected to the processor 840. Processor 840 is configured to sense an audible alarm sound in a sample of digitized sound from the A / D converter 830 in the manner described above. Similarly, the memory 850 and the power supply 860 perform the same functions as the memory 350 and the power supply 360 in FIG.

The smoke detector 800 further includes a smoke detector circuit 805 connected to the processor 840. The smoke sensing circuit may be any conventional type known in the art. Also connected to processor 840 is a transmitter / receiver 880 (typically an RF transmitter / receiver) capable of communicating with other smoke detectors (not shown in FIG. 8). When the processor 840 senses an audible alarm from another smoke detector (not shown in Figure 8), or when the smoke detector circuit 805 detects a fire, or from another smoke detector via the transmitter / receiver 880. When the operation is received, the processor 840 sounds an audible (and / or visible) alarm 895. In addition, processor 840 sends an activation message to the tactile alarm device 870 and an additional smoke detector 890.

The processor 840 of the smoke detector 800 activates the alarm devices 870 and 895 when it receives an activation message from another smoke detector via the transmitter / receiver 880, or when an audible alarm sound of another smoke detector is detected. Activate. In an alternative embodiment, both conditions must be met for the alarm devices 870 and 895 to operate. This helps reduce false alarms, but makes the sensor 800 more susceptible to RF, or auditory interference.

Those skilled in the art will appreciate that it is desirable to send an activation message as soon as it receives an activation message from another sensor. This is because it allows the detector 800 to act as a repeater when three or more detectors are installed in one place. Thus, for example, the first smoke detector may send an activation message that is received by the detector 800 but not by the third smoke detector. In this case, the activation message sent by the detector 800 immediately after receiving the activation message from the first detector may reach the third detector.

It should be understood that in some embodiments, the smoke detector 800 does not include all the components of FIG. For example, the smoke detector 800 is not used with the tactile alarm device 870 when it is not necessary to warn the hearing impaired. Such embodiments may be used to provide station-to-station operation without the need for wiring. Further, such an embodiment may not include a microphone 810, an amplifier 820 and an A / D converter 830 for sensing audible alarm sounds from other detectors, instead from other detectors. Operation can be triggered depending on the RF activation signal and smoke sensing circuit 805. As an alternative, other embodiments are not adapted to receive activators from other detectors via transmitter / receiver 880, instead audible alarm sounds from smoke sensing circuit 805 and / or other detectors. Triggers the audible alarm 895 and the tactile alarm 870, depending on the perception of.

The present invention has been described with reference to specific examples of smoke detectors and methods and devices for detecting the audible alarm sound of smoke detectors, but according to those skilled in the art, it deviates from the spirit of the present invention. It will be appreciated that many modifications and changes can be made without it. Therefore, the accompanying claims are intended to cover all such modifications and modifications within the true spirit and scope of the invention.

<figref num="1">It is a timing diagram which shows the time pattern of the audible smoke detection alarm sound by the NFPA standard.</figref><figref num="2a">FIG. 6 is a plot of the amplitude as a function of the frequency of a typical audible smoke-sensing alarm sound obtained by a sensor in a box with a foam lining.</figref><figref num="2b">FIG. 3 is a plot of the amplitude as a function of the frequency of a typical audible smoke-sensing alarm sound obtained by a sensor in an unlined box of foam.</figref><figref num="3">It is a block diagram of the hardware of the device which notifies the operation of the smoke alarm according to 1st Embodiment of this invention.</figref><figref num="4">It is a flow diagram of the acquisition routine executed by the device of FIG.</figref><figref num="5a">Together with Figure 5b, we construct a flow diagram of the sensing subroutine executed by the device of Figure 2.</figref><figref num="5b">Together with FIG. 5a, construct a flow diagram of the sensing subroutine executed by the device of FIG.</figref><figref num="6">It is a figure which shows the total sample period for the subroutine of FIG. 5a and FIG. 5b.</figref><figref num="7">It is a figure which shows the individual sample period of the subroutine of FIG. 5a and FIG. 5b.</figref><figref num="8">It is a block diagram of the hardware of the smoke detector according to the 2nd Embodiment of this invention.</figref>