GB2188812A - Telephone answering device - Google Patents

Abstract

An automatic telephone answering recording and playback device where the incoming and outgoing messages are digitised and stored in memory integrated circuits. A microprocessor controls the memory so that messages can be played out to a distant calling party, or a distant called party, and messages can be recorded from a distant calling party, under the manual control of the user, or controlled by a pre set programme, or by the detection of verbal signals using voice recognition techniques, or by a combination of the aforesaid. <IMAGE>

Description

SPECIFICATION Telephone answering device This invention relates to telephone answering machines which answer the telephone to a distant calling party, transmit an opening announcement to the distant calling party and record one or more incoming messages. It is an object of the invention to provide a telephone answering device as described which has no moving mechanical parts, except for switches, has neither a magnetic tape system for recording and playing back the opening announcement nor for recording and playing back the incoming messages.

According to the invention, voice signals in analogue form are translated into digital form by means of an analogue to digital converter. The signals are then stored into RAM, i.e. a random access memory integrated circuit. Digital voice signals are retrieved from the RAM and reformed into analogue signals by a voice synthesizer, then transmitted out to the telephone line or locally to a loudspeaker.

In one or the preferred embodiments of the invention, the common parts of the system are the analogue to digital converter and voice synthesizer circuits. Either separate random access memories or segments of a single random access memory hold, the digitized signals that is the outgoing messge, and the incoming messages, and these are played back either to line or to the local loudspeaker at selected or pre-determined time intervals initiated by a microprocessor.

In a second embodiment, after an incoming signal is recorded into the random access memory, it is followed by digital signals which correspond to the day time and day, so that when the user hears the playbackofthe recorded messages, it is immediately followed by the time day and date information in voice signals.

In a third embodiment, the system can be remotely controlled by the user. When the user speaks into a distant telephone using coded words, a voice recognition circuit will interpret the meaning of such words and control the answering machine accordingly. Such coded words word operate as follows: If the user says 'Playback' from a disant telephone, the system would playback all the recorded messages. If he says 'Repeat' the system would playback the messages again. If he says 'Record announcement' the user can record a fresh announcement from a distant telephone and the system will play back the recording. If the user says 'Save the messages' the device will disconnect itself from the telephone line, but all the recorded messages will remain stored in the random access memory.

If the user says 'Erase messages' all the recorded messages will be erased from the random access memory.

Atelephone answering device according to the invention will now be described by way of example with reference to the block diagram in Fig. 1, showing the operation of the electronics that form the device.

To record an opening announcement, a key on the key pad 5, marked 'opening announcement' is depressed, and a corresponding light is illuminated on the visual display 18.

The keypad 5 actuates the microprocessor 6, the analogue switch 2, the audio amplifier 3, the analogue to digital converter 4, the switch 8, and the RAM 7. When the user speaks into the microphone 1, the analogue voice signals are fed into the analogue switch 2, amplified by the audio amplifier 3, and converted into digital signals by 4. The digital switch 8 feeds these signals into RAM 7. After a fixed time, say 15 seconds, 6, will extinguish the appropriate lamp on 18, indicating that he has about 5 more seconds of recording time to go before the allotted time for recording the announcement is at an end.

Immediately, microprocessor 6, will turn off the amplifier 3 and switch 2, so that there is no input to the switch 8, from the converter 4. But 6, switches on 15, the output of the RAM and supplies the necessary clock and other signals to RAM 7 via 8, so that digital signals corresponding to the opening announcement and stored in ram 7, will be played out to voice synthesizer 19. The restructed analogue voice signals are amplified by audio amplifier 20 and conveyed to the loudspeaker 26, via the power amplifier 22 and switch 23. Switch 23 controlled by 6.

The for the device to be set to answer the incoming telephone calls, the used depresses a key marked 'answer' on the keypad 5. A lamp on 18, lights, indicating that the device is 'set to answer'.

The microprocessor, sensor the 'set to answer' command, allows the following sequence of events to take place.

When there is an incoming ringing signal on the telephone line 27, usually with a frequency of 10-40 hz, 26, detects the ring signal and commands 6, to answer the incoming ring. 21 is therefore activated and 21 places a dc loop across the telephone 27, this answers the incoming call.

The microprocess 6, the switches the system to play the outgoing messages to line. Switch 15 is activated by 6, and the voice output to line is conducted through,19,20,21.

The end of the announcement is detected by port 28 or 6, the and system then switches to record incoming messages by 6. It is not necessary for the system to transmit a beep tone to line to indicate to the distant caller to speak and leave a message, the user might record at the end of his opening announcement, 'Please speak now' and at the end of 'Now' the machine will switch to record. For detection of this the user simply depresses an end of recording button on key pad 5, so that the last word 'now' is registered in RAM 7.

During the record incoming message phase, port 32, of 6, switches off the audio amplifier and voice synthesizer through ports 30 and 31.

The output from the telephone line 27 is conducted switch 2, 6 switches 2, so that port 33, conducts the line signal to the audio amplifier 3, converter 4, and 6 seiects the appropriate ram and in this case for the first message, suppose the digital contents of the first message is dumped into ram 8.

Port 36 of 6, switches 2, at port 34 for recording incoming messages. An internal timer in 6, preset to say 30 seconds, will inform the distant calling party that his recording time is up as follows: 6, will switch the system in the playback to line mode as already described for the outgoing announcement.

A pre recorded sentence, digitized in ram 7, will hold the sentence 'Thank you for calling' or any end of call sentence that is recorded by the user using the keypad 5, and facilitities similar to the way he recorded the opening announcement.

At the end of the recorded incoming message sequence, 6, will activate 21, and disconnect the loop across the telephone phone and setting the device to answer further incoming calls.

At this point in time, a binary signal from the clock 16, synchronous with the mains frequency of 50 hz, is fed into ram 12 and stored and codes such that when on playback so the contents of ram 8 is delivered to the speaker 26b, 6 senses the end of the message sequence, delivers the appropriate command to 12, via switch 13, the matching binary data, will then produce digital signals from 12, through switch 14 to speaker 26b via 19 and 22 so thatthe listener will hear a message, say from 8, followed by his voice saying the time when the message was stored. Again the appropriate switching and sensing for this is accomplished by microprocessor6.

If the distant calling party telephone into his own machine he can ask the machine to playback recorded messages as previously explained. His voice signals, will be conducted through 21, and switch 24, switch on at the appropriate time by 6. His signals are detected by 25 and the command to 6 is given through port 32.

Incoming messages are stored in ram 8, 9, 10 . . . 11 or as many rams as desired and selectively played back to the user, either remotely as just described, or locally, as follows. 6 will sequentially select 8,9, 10 etc. and switch the appropriate digital signals through 15 and the voice synthesizer 19, audio amplifier 20, switch 23 the audio power amplifier 22 and speaker 26.

The advantages of this device are as follows: no moving parts and thus reliable operation. A common voice output synthesizer circuit 19,20 for all levels of operation, a common voice input circuit 2,3,4 for all levels of operation. The recording capacity is limited by the capacity of the memory units, 7, for the outgoing announcement and 8,9, 10 etc. for the incoming messages. For example if the capacity of a particular ram say 8 is for 15 seconds of speech, this can be increased by adding equivalent rams in parallel to increase the capacity, so that that two 8's, in parallel would accommodate 30 seconds of speech and soon.The memory in the microprocess 6, would be programmed to parallel segments of the ram to suit the timing of incoming calls, i.e. voice operated calls, or the microprocessor could be set to a fix time iimit, so that only x seconds of memory could be utilised and the distant calling party would be signalled that that was the end of the message sequence.

In a fourth embodiment the system can be made to change the announcement remotely. After the machine has answered the distant calling party by placing a loop, i.e. a dc loop across the telephone 27, as described above, and after the announcement has been played to line, the distant calling says into the distant telephone 'Change Announcement' of Change G reeting' or words to that effect, according to the pre programming of the voice recognition circuit, 25 which will be described later. The system will immediately change to record announcement, as described earlier, but in piace of the microphone connection 1, to switch 2, 21 the line interface is connected to analogue converter 4, via 2, 3.As described before after the announcement has been recorded and after a fixed time of about 15 seconds, the system will immediately go into playback and back the announcement to the distant calling party for confirmation. If he hangs up after the playback, the announcement is retained. If he wants to change the announcement yet again, he simply says 'change announcement' after the playback and records a new announcement. A beep tone transmitted from 6, will inform him 5 seconds before the end of the 15 seconds announcement period.

In a fifth embodiment, the system can be made to record a personal announcement by saying into the distant telephone 'Record Personal Announcement' Similarly the system will playback a personal announcement if the distant calling party speaks into the distant phone 'playback personal announcement'. This personally announcement might be understood to mean also a personal message.

In a six embodiment a personal message may be recorded locally, and retrieved remotely as indicated above. To record a personal announcement locally, the user, depresses a pre arranged combination of keys on the key pad 5, and the system will record a personal announcement via the microphone 1, to a ram, say part of 7, but identified as a personal announcement as the coded voice signals will be preceded by a binary code, delivered by 5,6,8 and thus to 7.

In a seventh embodiment the machine can playback messages as they are received to another telephone number or alternative telephone number.

This function can be programmed locally via the keypad 5, or remotely, via the voice recognition circuit 25. To program locally, the used dials in a pre arranged code on the key pad 5, he might key in a pre arranged digital code, say 123, then followed by the telephone number say, 01-999-9999. The code 123 would be compared with the redialling code, programmed in the memory of the microprocessor 6.So that after a message has been received, 6 will, through its port 411 activate 21 into the dialling out mode via 41, and 6 via 411 would dial out the programmed numberto line 27. 6, through port412, and audio line sense amplifiers 43, will sense that the distant receiving party at the other end of the telephone line has picked up his phone after answer the outgoing call, this is sensed by 43, by 'listening out' for end of ringing tone and transmitting the last message received by ram 8, or 9 or 10, or 11 etc.

Again this accomplished by 6 selecting the last message received or stored in the ram and switching the system so that the message is played out to line, as previously described. This seventh embodiment may be activated remotely or locally by using the voice recognition circuit 25, in this case, user instead of dialing a pre arranged digital code followed by the telephone number as described in this embodiment, the user, says either into the microphone 1, or distant telephone connected to line 27, 'send all messages to 01-999,9999' For confirmation, the system will playback to line to to the loudspeaker,26, for the userto hear 'I will send back all messages to 01, 999,9999 until further commanded'. . . or words to that effect.Messages may be restricted by commands such as 'send Personal messages only to 01,999,9999' Or by the command 'send only last message received to 01,999,9999.' The system may be restricted further, so that if the distant calling party says a pre arranged coded word or sentence, when the machine in is in the recorded message mode, the machine will play back messages only to those persons saying the correct coded, words or phrases.

The voice recognition circuit 25, is an input output device and an example of such a system is described in Electronics, 1983, April 21, pages 128 to 143.

Circuit 25 envelops four basic steps, voice feature extraction, voice pattern similarity measurement, time domain registration and a decision strategy programmed into a memory system.

In addition the circuit 25 must 'learn' and retain in its memory speaker dependent reference patterns which key the recognition to individual speakers.

Because of the wide variations among different speakers, speech recognition systems must be tuned or enrolled to accept speech from only one or perhaps a limited number of users.

The system of enrollment creates a set of feature vectors or templates for each word or command in the systems memory. The templates which are included in the similarity measurement as part of the recognition process define the spectral shape of the reference pattern for individual words held in the reference memory. All individual and multiple word utterances of the overall vocabulary must be enrol led. Successive repetition of the utterances leads to significant immprovements in performance of the system as a recognizerofthe individual speaking. The substitution error rate decreases substantially with each individual training session by the user. After 5 training inputs for example, substitutions generally decrease to about one third the rate that exists after the first training pass. See Figs. 1,2.

Training is accomplished by the user pressing a pre arranged code on the key pad 5, this code will indicate to the microprocess 6 and 25 that a training session is about to begin. After pressing the GO key (not shown) on 5, the user says one of the many commands described previously i.e. 'Erase Messages' The user should press 'Go' and repeat 'Erase Messages' he should press 'GO' again and repeat the phrase. He should continue to do this for 5 or 10 times.He should then depress 'Test' button on the keypad 5 (this but is not shown in the drawings) He should then say immediately into the microphone 'Erase Messages' Immediately after, he will hear through the speaker the words in his own voice 'Erase messages' which would confirm that the system understands his voice and phrase for this event or he would hear 'Error' in this case he must press the 'GO' button on key pad 5 and retrain the machine with more 'Erase messages'.

In order for the system to recognise the phrase 'erase messages' from a distant telephone iine, the voice input to 25, from the microphone 1, and contained within 25, are filters which transform the voice signals from 1, into the voice quality one might expect when the voice is transmitted by telephone, across short lines or long lines to the device, via telephone exchanges and repeaters.

After each training utterance, filters in 25, are sequentially inserted between the input from the microphone and the input of the pattern similarity measurement circuits, these filters would represent various distortions to the human voice over different line impedances, line lengths, and distortions to the original waveform which is to be expected when voice is transmitted over telephone lines, and via various microphones inserted in various telephones, i.e. carbon microphones, electret microphones etc.

Feature extractions from the original training waveforms transforms the speech signals into time varying parameters that can be recognised by part of the circuitry of 25.25 will also compute the measured similarity between the extracted speech signals from the telephone line and the local microphone with the stored reference patterns.

When the reference data matches the input data, i.e.

when the spectral peaks and valleys of the waveform of the reference data match the spectral peaks and valleys of the input data, a residual error signal will occur. This error signal is measured and computed in circuit 101. If there is a perfect fit between input signal and signal in the memory then there would be zero error signal.

To be usable the error signal should differ from the matched pair by less than 5%.

The parameters that are used to compute the difference between the input samples and stored reference patterns are computed by autocorrelation algorithms 102 and residual energy algorithms, 101.

The time registration function, 103, allows recognition of connected speech and is performed by a type of dynamic programming that compensates for variations in the length and timing of the input utterances, without changing the meaning of the inputs. 105, enrolled vocabulary memory. 106, event detector, 107, logic 108, host computer, 109, multiplexer, 110, ram, 111,auto address register, 112, buffer, 113, analogue to digital encoder, decoder, 114, analog, input, output, program ram 115, 116 Texas Instruments 320 digital signal processor.

It should be understood that the voice recognition system in this invention can be replaced with a simple conventional tone decoder system, whereby the distant calling party carries a 'beeper' which emits an audible tone when activated. The beeper is held againstthetelephone microphone, the user activates the beeper and the signal is detected by the system f shown in Fig. 1. 25, is replaced by a tone detecting circuit.

The system then switches, remotely, by the user into various play and rewind modes. The method of recording storing and playing back the speech signals is the same as described.

References: Voice recognition and synthesis device: General Instrument, Integrated circuit No.

SP 1000. SP 1000 specification includes, recognition features, synthesizer features, architechrual description, operation codes.

'Electronics' is published by McGraw Hill in the US.

An example of converting analogue sound signals into digital form using an analogue to digital converter (this would represent 3,4 in Fig. 1, may be found in 'Simple Interfacing Projects. Bishop. pages 116-135. Published by Granada 1983.

Examples of telephone answering machines, but using magnetic tapes for recording voice information: Rubenstein UK patent No.1,193,081.

Rubenstein US Patent 4006312.

Examples of Drams and Srams shown in Fig. 1,8, 9, 10, 11 are in Microelectronic Data Book, Published by MostekCorp, 1984, U.S. Publication. MK 4856, 32Kx8 bit dynamic ram, MK 6116 may be quoted as an example of a CMOS SRAM.

However, to attain about 15 seconds speech to be held in the memory of rams shown in Fig. 1,8,9, 10, 11. Rams of the order of 256K 8 bit, would be required. A256K DRAM is manufactured by Hitachi in Japan.

It should be understood that Rams 8-11 in Fig. 1 can be extended in number to increase speech storage capacity. By way of example a bank of 20 DRAMS i.e. adding to the rams 8-11 could provide speech storage and playback capacity of 5 minutes of recording and playback, is of the type that contains a CPU, on chip clock, ROM, RAM, I/), timer.

One example of this type although the illustration shows only a limited memory capacity is the S 6805 of American Microsystems Inc and described in their catalogue 1984, pp 6.14-6.31. This invention requires ROM and RAM capacity larger than shown in the S 6805.

The keyboard, 5 is of the type used on push button telephones and may consist of up to 30 push buttons.

Claims (3)

1. A telephone answering device comprising means to digitise the analogue signals of outgoing and incoming messages, means to store in memory integrated circuits said messages, means to retrieve said messages, from said memory integrated circuits, means to convert said memory into analogue form for transmission to loudspeaker or telephone line, voice recognition means whereby the device is controlled by the voice of a distant calling party.

2. As in claim 1, whereby the device is programmed by the voice of the distant calling party using voice recognition techniques so that the device dials out to a distant receiving party.

3. As in claim 1 whereby the device is controlled by a distant calling party using audible tones.