This invention relates to a responder designed to detect and respond to a primary alarm signal (eg the sound of a smoke alarm) by producing a secondary alarm signal.

Responders of the abovementioned type have been proposed in the past eg to give a tactile alarm signal to people with impaired hearing, or to illuminate an escape route from a burning building or a route through which rescuers are encouraged to enter a building. In all of these systems, it is important that the secondary alarm signal should start as soon as possible after the first so that there is no significant delay. One known system is described in patent specification US5177461. This describes a gadget that is designed to be placed on a window of a room containing a smoke alarm. An output from a microphone in the gadget is filtered, rectified and applied to a low pass filter. The resulting DC signal is used to trigger a switch to activate a strobe light, directing rescuers to enter the building via the abovementioned window, directly into the room where the fire has been detected.

It is worth mentioning that the abovementioned proposal would be recognised by professional fire fighters as dangerous because, trying to effect a rescue by opening or breaking a window is likely to increase ventilation and worsen the fire. The circuitry is also likely to create many false alarms because it could respond to noises from many different sources such as telephones, bird song, music etc.

Another known proposal is described in patent specification US7015807 and employs a processor to perform a Fourier analysis of detected sound signals and to compare the results with what would be expected from a smoke alarm. This technique has the problem that a processor capable of performing Fourier analysis on an incoming audio signal consumes considerable power. Reliable functioning of the device also relies on a prior knowledge of the characteristics of the smoke alarm with which it is required to co-operate. This latter point is important because smoke alarms in most countries have no standard sound pattern. The invention provides a responder designed to detect and respond to an alarm signal consisting of bursts of energy comprising: an energy sensor and a processor connected to receive signals derived from the energy sensor, the processor being programmed to provide means to detect a degree of consistency in one or more characteristics of the bursts and to activate an indicator in response to such detection

.By employing the invention it becomes possible to eliminate the problems associated with high false alarm rates, high processor power requirements and the dependency on a knowledge of the characteristics of the smoke alarm smoke alarm. Furthermore, because the processor needs only to detect consistency in the pattern of on/off pulses in the received signal, the processing power is a magnitude less than would be necessary for a Fourier analysis of the signal. At the same time, the system can be made to respond to signals from many different smoke alarms because they all generate a pattern of pulses that is consistent over a period of seconds or minutes even though the patterns themselves and the frequencies vary greatly. The responder is preferably designed to detect sound energy from standard smoke alarms but would also be applicable in a system where a smoke alarm is adapted to broadcast a coded ultrasonic or wireless signal that is received and decoded in the responder.

The programming of the processor is preferably such as to examine a sample of the signal for consistency over a period of 15 seconds or more preferably 30 seconds. This should be sufficient to eliminate responses from most extraneous sources such as bird song or music. Even better results are obtained with periods of 1, 2 or even 3 or 4 minutes and this will allow sufficient time for the smoke alarm to be reset after a minor incidents, that is not a true emergency, eg burning toast, to be dealt with. A period in excess of 4 minutes is not considered satisfactory since that might result in the secondary alarm being activated after rescuers have arrived. The UK Fire and Rescue Service (which is typical of similar services worldwide) estimates that the minimum response time between receipt of a call for help and arrival at the scene of a fire is 4 minutes in an average urban location, though in remote locations it may be very much longer. The responder preferably includes a wake-up circuit connected to receive an output of the sensor (normally a microphone) and effective to produce a wake-up signal identifying the continuing presence of the said bursts over a period of time; the processor being arranged to be woken from a dormant state by the output of the wake-up circuit. The use of the wake-up circuit, which may easily be provided by a rectifier followed by a low pass filter, means that the processor needs to consume power only on occasions when a sound possibly, from the primary alarm, has been detected.

The processor may examine any property of the signal for consistency. Possibilities include i) the frequency of an audio signal over a number of bursts ii) the period of the bursts ie the time between equivalent points in adjacent bursts and/or iii) the on/off ratio of the bursts. In a preferred arrangement all three of these properties are examined.

One way in which the invention may be performed will now be described by way of example with reference to the accompanying drawings in which: -

Fig 1 is a schematic plan view of a building fitted with a system in accordance with the invention;

Fig 2 is a circuit diagram of an indicator unit mounted outside one of the rooms of the building; and

Fig 3 is a flow chart showing the principal operations performed by a processor indicted on Fig 2.

The "indicator" that is activated in response to the alarm signal could be a secondary alarm intended simply to reinforce the effects of the first-mentioned alarm. However, the invention is of special significance where the "indicator" is a light source positioned to be visible from an access area in a building for the purpose of guiding rescuers to a room known to be occupied by a vulnerable person such as a child or disabled person. In suchan arrangement, any delay, of up to 4 minutes, caused by the need to examine the signal for consistency and caused by the wake-up operation (if included) is of no disadvantage because it is known that rescue crew cannot ever be expected to arrive at the scene of an emergency in less than 4 minutes from a call for assistance.

Referring firstly to Fig 1 there is shown a building having an outer wall 1 , a number of rooms IA to IH and an access area A. Doors 3 open into rooms IA to IH respectively and a door 4, in the outer wall, opens into the access area A. Each of the rooms IA to IH and the access area contains a ceiling-mounted smoke detector SD of conventional construction. This is designed to emit an alarm signal in the form of a sound burst having a frequency of f and a duration of ti. This sound burst is repeated indefinitely with periods of to between bursts. The values of f, ti and to vary depending on manufacture. They generally use a base frequency of 2.5KHz to 3.8

KHz. Pulse rates vary from 1 to 4 Hz and the on/off ratio from 95% to 2%.

In Fig 1 it is assumed that room IF has been identified as being occupied by a vulnerable person. Adjacent to the door 3 of this room and visible to persons in the access area A is a visual signalling unit 5. This is designed to respond to an alarm signal from any of the smoke alarms by emitting a flashing light to direct a rescuer, on entering the access area A, directly to the room containing the most vulnerable person.

Referring now to Fig 2 there is shown a microphone 6 comprising an off-the shelf piezoelectric device 6A, adjustable resistor 6B, capacitor 6C and resistor 6D.

The output of the microphone 6 is passed to an amplifier and filter 7 having three identical stages 7A, 7B and 7C. Each of these stages includes an OP amp such as component MCP6144 available from Microchip Technology Inc. used in a multiple feedback configuration and designed to pass frequencies within the band 2.5 to 3.8 KHz. A pair of diodes 7D between the second and third stages prevents the third stage from saturating and causing excessive demands of the battery.

The output of the amplifier 7 is passed to a square wave generator 8 comprising a capacitor 8A and a Schmitt trigger 8B. The resulting square waves, still at audio frequencies, are passed to a programmed microprocessor 9 such as component PIC18F24K20 of Microchip Technology Inc.

The output of the amplifier 7 is also passed to a wake-up circuit 10 where it is first rectified by an arrangement of diodes 1OA and capacitors 1OB and resistor 1OC. The rectified signal is then passed to a low pass filter formed by resistor 1OD and capacitor 1OE. When the rectified voltage, has exceeded a certain level, say 2 volts, for a significant period, say 1 second, the voltage on capacitor 1OE becomes sufficient to operate a Schmitt trigger 1OF, indicating the presence for that 1 second period, of a sound having a frequency consistent with that of a smoke detector. The Schmitt trigger, when triggered, switches and holds on the power from battery Bl to the microprocessor 9. When there is no detected sound, the microprocessor is switched off by the circuit 10 thereby avoiding demand on a battery Bl. The battery supplies power to each of the components 7, 8, 9 and 10 but the demand from the microprocessor 8 is the greatest.

In addition to receiving timing signals from the square wave generator 8, the microprocessor has access to a store 11 containing, for all known commercially available smoke alarms, information defining i) the range of average audio frequencies, ii) the range of periods between the start of any one burst and the start of the next burst, and iii) the range of on/off ratios. It is programmed to perform a logical operation which will be described later with reference to Fig 3 and to produce a pulsing output (of 0.5 Hz with an on/of ratio of 2/7) only when it has detected a consistent pattern of input signals for a period of one minute.

The output from the microprocessor is fed to an LED driver circuit 12 powered, in this particular example, from a separate battery B2, to drive an LED 12A. After a rescue operation, the system is re-set using a reset switch RS connected to the processor.

Referring now to Fig 3, when the processor 9 has been switched on by the wake-up circuit 10, a real time clock 13 generates interrupt signals at times T ₁ , T ₂ , T3 etc, defϊning time slots between them. In this example, each time slot is 20 ms. The square wave generator 8 of Fig 2, also shown on Fig 3, produces bursts of square waves corresponding to the bursts of noise detected from the smoke alarm that has activated the wake-up circuit. During each time slot, the zero crossing points (two for each cycle of the square waves) are counted and stored at 14. At 15, the count for each time slot, representing the instantaneous frequency of the input signal, is compared with a set range of counts which are known to be generated if the detect sound has emanated from commercially available smoke alarms. In this example, the set range of counts is 100 to 200, corresponding to smoke alarm frequencies of 2 to 4 KHz.

If the incremental count derived at 15 is outside the set range, it is assumed that the current burst of sound or "bleep" has finished and the finishing time, eg T _n is recorded at 16. When an incremental count within the set range is next detected, it is assumed that a new burst of sound has started and the start time is, eg Tn+ 1, is recorded at 17. The count values from 14 and the burst start and ending times from 17 and 16 respectively, are processed at 18 to derive, for each burst and subsequent period of silence: i) the average frequency of the burst, ii) the period of the burst, and iii) the on/off ratio. These values are all compared at 19 with ranges of corresponding values for known smoke alarms stored at 10. These ranges of values need to be wide because of the wide differences between different alarms. For example value i) may be specified as being within a range of 2 to 4 KHz; value ii) between 0.5 to 2 seconds; and value iii) 95% to 50%. If no match is found the power to the processor 8 is switched off.

In an alternative arrangement, the store 10 contains specified values i), ii) and iii) for every known make and model of smoke alarm. Although it is still necessary to have a margin for error in the comparison process, this margin does not have to be as wide as the ranges used to cover all alarms. However, a problem with that alternative is that resetting of the store would be needed whenever a new alarm is marketed. If a match is found, a step is performed at 20 to test for consistency of the values i), ii) and iii) over the time since the wake-up circuit 9 switched on the processor. If the values are all found to match within a specified tolerance (in this example 10%), a score, held in a register 21 is incremented by 1. If any one of these values are found to have varied by more than the specified tolerance, the register is decremented by an amount greater than 1. In this particular example it is decremented by 3. In this way, the register 21 contains a measure of the consistency of the sensed sound. This is important because, although different smoke alarms have a wide variety of different sounds depending on make and model, they can all be relied upon to be highly consistent over a long period of time.

If the score held in register 21 reaches zero (it is not permitted to be less than zero), this is detected at 22 and the power to the processor 9 is switched off. If the score reaches a value corresponding to a 2 minutes of processor switch-on time (in alternative embodiments this could be up to 4 minutes), this is detected at 23, causing a strobe circuit 42 to produce a pulsed output to the LED driver circuit 11 before the fire crew arrive at the scene of the fire.

The embodiment of the invention that has been described is particularly effective and reliable because it usefully employs, for processing, at least a substantial proportion of the minimum time taken for a fire crew to respond and because it examines the signal for consistency of the parameters referred to by references i), ii) and iii) above, ie it measures changes in properties of the signal rather than the more obvious choice of relying on measurements of the absolute values of those properties, those values, in any event, being of uncertain usefulness because of the wide variety of different sounds emitted by smoke alarms.. Many variations to the described arrangement are possible within the scope of the invention as defined by the accompanying Claims. For example, the store 10 could be loaded with values i), ii) and iii) for every known make and model of smoke alarm. Although it would still necessary to have a margin for error in the comparison process 37, this margin does not have to be as wide as the ranges used to cover all alarms. However, a problem with that alternative is that resetting of the store would be needed whenever a new alarm is marketed.

In another variation, the one minute time period during which processing takes place and the one second timing period of the wake-up circuit 9 could be varied. Another possible variation would be to arrange the microprocessor to switch on the amplifier 6 only intermittently. In theory, it would need to be switched on for just one second (sufficient for the wake-up circuit to respond) each 4 minutes (the minimum time taken for a fire rescue crew to arrive, though in practice some good margins for error are desirable.

Finally, it should be mentioned that, although the invention is thought to be most likely to be employed in relation to fire emergencies, it is equally applicable to emergencies associated with dangerous heat levels (without combustion) and emergencies associated with dangerous gasses.