Service node for a telephony network

2~95~37 A24679 (P05) SERVICE NODE FOR A TELEPHONY NETWORK The present invention relates to a service node connectable to a telephony network, in which a plurality of peripheral devices are provided, along with switching means for connecting said peripheral devices to external communicationmeans under the control of control processing means.

Over recent years, telephony networks have become more sophisticated 10 and it has become possible for an increasing number of services to be provided, in addition to the fundamental provision of allowing a first customer to be connected to a second customer.

Telephony systems of this type are often referred to as intelligent networks, and provide such services as automatic call distribution, store and 15 forward, fax to speech, call queuing, voice menu systems, call diversion, follow-on and voice prompts, etc.

Service nodes are units within the network where intelligent services are concentrated and made available to customers connected to the network via switching mechanisms provided within the network.

At the node itself, a large 20 switching device is provided which allows customer calls to be connected to peripheral devices under the control of central control processing means, such as a mini or mainframe computer.

In known systems, an incoming call is connected to a digit collector to collect, or receive, digits (i.e. signalling information) generated by a calling customer, in order that peripheral devices may be connected and, as25 necessary, disconnected etc.

Thus, once a particular service has been established, this is generally driven under the control of a particular peripheral device.

However, it is necessary for the central control computer to observe the operation of a connection between peripheral devices and external lines, in order to ensure that the integrity of the overall network is maintained.

A problem with known service nodes is that it is desirable to have large switches at the nodes, so that maximum benefit may be gained from a large number of shared services.

However, as the number of shared services increases, with a resulting increase in the size of the connecting switch, the size and power AMENDED SI~ ~19543~ Z ~C~ of the control computer must also increase, along with the bandwidth of the communicating link between the computer and the switch.

It will be appreciated that the use of dual tone multi-frequency (DTMF) receivers to receive and decode routing digits or other command signals is well 5 known in the art as is exemplified by the following prior published documents.US Patent No. 4,879,743 discloses a PBX system which is responsive to an incoming call directed to an adjunct of the system, for example a voice messaging system, for sending a ringing signal to the adjunct.

When the adjunct goes off-hook, the PBX outputs to the adjunct dual tone multi-frequency (DTMF) 10 signals containing information associated with the incoming call prior to establishing a connection between the adjunct and the incoming call.

The information includes a mode code for selecting the required operating mode of the adjunct, and also includes the calling and called numbers.

Following the sending of this information to the adjunct, the PBX then sends a disconnect mode code to 15 command the adjunct to go on-hook, and subsequently connects the adjunct and the incoming call.

During this connection, the adjunct may use in a well-known manner prerecorded voice response to prompt the calling party to enter DTMF digits to access a particular operating capability of the adjunct.

In this US Patent No. 4,879,743, the adjunct, which is a program20 controlled device, has to identify which of the set of mode codes of the form##nn, where n is a numerical digit in the range 0 to 9, has been received, and, in response to a particular mode code identification, switch to the corresponding operational mode.

It does not communicate the received mode code to the PBX or any other adjunct or terminal equipment attached to the PBX.

US Patent No. 5,181,283 discloses an arrangement for providing preauthenticated access from a caller to a service provider.

The caller dials the number of the service provider, and his local switching system routes the call to an originating service office (OSOI.

The dialled number is passed by the OS0 to a network control point (NCP) which translates the dialled number to a non-dialable 30 directory number (NDDN) together with the identity of an action control point(ACP) having as appropriately equipped associated network services complex (NSC) for routing the call to the destination service provider terminal, and returns this information to the OS0.

The OSO forwards this information to the ASP which AMENDE~ SHEE~ 2 1 9 5 4 3 7 ;;

.. .. .. ..

queries a direct services dialling data base network control point (DSD NCP).

The DSD NCP returns a message to the ACP a voice announcement message identification and a reply format indentification including the number of digitsexpected from the customer.

This information is the passed to the NSC which 5 conducts the subsequent exchange with the customer.

The NSC detects the DTMF digits keyed by the customer in response to a prompt, these being a personal identification number (PIN) and, if the customer's account number has not been obtained using his originating telephone number and a lookup table, an account number.

The NSC sends to the DSD NCP a message including the 10 customer's account number and the NDDN.

The DSD NCP checks that the desired service provider has an individual customer account file corresponding to the supplied account number, and provides to the NSC the PIN corresponding to the suuplied account number.

Upon receipt of the requested information, the NSC verifies that the caller has been authorised to access the service provider, by 15 comparing the PIN provided by the customer with the PIN provided by the DSD NCP, and sending a message to the ACP to forward the call request together with the caller's account number to the service provider.

There is no disclosure of digit collection for any purpose other than establishing an authorised call, involvingrouting digits, account number digits, and PIN digits.

The article "The AT&T Service Circuit Node: A new element for providing Intelligent Network services" by Hall et al, AT&T Technical Journal, vol. 70, no.

3/4, 1991 Short Hills US, pages 72-84, discloses a stand-alone service-circuit node that connects to AT&T's 5ESS switching centres.

The node comprises a control computer; a number of switch-fabric units; and a collection of service circuits 25 which provide such capabilities as voice announcements, DTMF receivers, and text-to-speech conversion units.

When a call is directed to the node from a 5ESSoffice, in response to recognition by the office that a special service is required because the dialled digits (control information) do not represent a network destination for which the office can proceed to effect a connection, the control30 information is sent to the control computer, which determines the service to be performed and instructs the switch-fabric unit to connect the call to one or more service circuits (sequentially) to collect digits, play announcements, receive and record a fax transmission, play a message through a text-to-speech unit, etc.

The A~IIENDED StlEEY 2 1 ~5437 ~ ~.~ article merely mentions that an incoming call can be connected to a digit collecting service circuit.

The conventional use of received digits in these circumstances would be as a response to a voice prompt (announcement), e.g. to collect an account number or PIN, or to make a selection from an offered menu.

According to a first aspect of the present invention, there is provided, a service node for use in a telephony network, the service node comprising a plurality of peripheral devices;

switching means for connecting said peripheral devices to external communication means;

detecting means associated with said switching means and arranged to detect signalling information in communications channels incoming to the servicenode from said external communication means; and control processing means, the service node being characterised in that said detecting means is formed by distributed processing means and is 15 arranged to detect respective signalling information in each incoming communications channel, said control processing means is arranged to generate a representation of a scan pattern of signals and to supply to said distributed processing means said representation and the identity of a said incoming communications channel in 20 which said pattern of signals is to be detected, and said distributed processing means is arranged to perform scan recognition, in accordance with said received pattern, upon detected signalling information in the identified incoming communications channel and, upon recognition, to supply the detected signalling information to said control processing means.

Thus, distributed processing means are provided within the switch itself such that, rather than supplying to the control processing means all signalling information detected in calling customers' speech channels, the only informationwhich is sent to the control processing means is that which matches a scan pattern provided by the control processing means to the distributed processing 30 means.

In this way, for a given size of switch, demands made upon the controlprocessing means are reduced and the degree of bandwidth between the control processing means and the switch is significantly reduced.

For example, in the AMENDED SHEET 21 95437 . '; ~ ~ 3 preferred embodiment, a central control computer communicates via a standard ethernet connection.

In a preferred embodiment, the distributed processing means includes a first type of processing means (which may be digital signal processors) arranged to 5 produce an indication of detected signals; and a second type of processing means (which may be a microprocessor) arranged to receive indications from said first type of processing means.

Preferably, digital signal processors detect dual tone multifrequency signals and convert these into information in the form of digital representations of 10 the conveyed codes.

Furthermore, in a preferred embodiment, distributed microprocessors are arranged to compare said detected codes against predetermined pattern permutations.

According to a second aspect of the invention, there is provided a method of recognising predetermined externally originating signalling information at a 15 service node connectable to a telephony network, comprising steps of :defining, on a control processing means, a scan pattern of signals to bedetected in an incoming communications channel;

downloading, to distributed processing means distributed in a switching means, said scan pattern and the identity of a said incoming communications 20 channel in which said pattern of signals is to be detected;

detecting signalling information in the identified incoming communications channel;

performing scan recognition, in accordance with said received pattern, upon the detected signalling information in the identified incoming communications 25 channel; and upon recognition, supplying the detected signalling information to said control processing means.

The invention will now be described by way of example only, with reference to the accompanying drawings, in which:

Figure 1 shows a public switched telephony network, including a plurality of customer terminal equipment, local exchanges, trunk exchanges and a service node;

AMENDED SHEET 2195437 ~ 3a.

Figure 2 shows the service node identified in Figure 1, including a control computer, peripheral devices and a switch for connecting external lines to said peripheral devices;

Figure 3 details the switch shown in Figure 2, including a switching matrix 5 and a plurality of line interface cards;

Figure 4 details a line interface card shown in Figure 3, including line interface circuits, digital signal processors and a cross point switch for connecting line interface circuits to digital signal processors;

AMENDED SHEE~ W096/04756 ~ 9 5 4 3 ~ PCT/GB95/01823 Figure 5 details operations performed on the line interface card shown in Figure 4; and, ~igure 6 detaiis control operations for the service noae.

~ telephony network is illustrated in Figure l, including a plurality OI customer terminal equipment ll, inciuding telephones and fax machines etc.

Each customer terminal equipment ll is connected via a local line 12 to a _ocal exchange 14, thereby providing by-directional communication between the terminal equipment and the local excAanges.

Local exchanges 14 are connected to trunk exchanges l5 and bi-directional communication between said exchanges is effected over trunk cables 16.

Similarly, trunk excnanges are bi-directionally connected by trunk iines l , again arranged to convey multiplexed signals.

In the preferred network, conversion between analogue sianals and digital signals occurs at the local exchanges l~ and communication throughout the trunk network is ~erformed in the digital domain.

Thus, each trunk line 16 ncludes physical communication links each arranged to convey 30 speecA cnannels over a 2 megabit per second mui-iplex.

In aadition, higher levels of multiplexing may be achieved using optical links, which are particularly suitable for connecting trunk ~ch~nges 15.

In addition to allowing conventional telephone connections to be made, the networR shown in Figure l also ir.ciudes a plurality of added value ser~ices which allow customers to interact via voice menus, receive recorded information from a central source, record messages and receive recorded messages, and establish sophisticated call diversion procedures, wherein, over a day, calls may be diverted to a plurality of numbers, reflecting the location of an actual customer, rather than the terminal equlpment.

These added value services are located at a central position, in the form of a ser~ice node 18.

The service node 18 is connected to the rest of the network W096/04756 21 95437 PCT/GB95/018~ via a plurality of 2 megabit per second digital links, sucA that any customer connected to the network may obtain ser~ices via the central service node by establishing an ap~ropriate connection through the network of exchanges.

The service node 18 is detailed in Figure 2 and comprises a plurality of peripheral devices; three of wAic;~ are identified by references 21, 22 and 23.

Each perlpheral device is configured to perform a specific tasK.

Thus, voice menu services may be provided by a speech applications platform, allowing two-way communication using -ecorded voice messages, voice reccgnition and DTMF tone recognition.

Another peripheral device may be arranged to supply recorded information to calling customers. A message is recorded centrally on the peripheral and written to -~nnomly accessible memory storage devices.

Thus, the same message may be played to a plurality o~ _alling customers and by addressing the memory devices inaividually for each call, the message may be played to each calling customer from the beginning of the message, c-eating the impression that a dedicated message has been recorded for that call.

.~ third peripheral device may be arranged to rec~rd messages left by customers so that they may then be collected by a specific customer for whom the service has been created, effectively providing the customer with the services of a virtual call answering machine.

Thus, on receiv'ing instructions from the customer, all of said customers will be directed to the ser~ice node such that, under the control of a control computer 24, voice data wil' be supplied to, say, peripheral 23, whereafter, in res~onse to a call made by the customer, said recorded information may be replayed.

~nother service provided by service nodes is that of redirecting calls. ~hus, all calls to a particular customer may be directed to the service node and, under the control of the control computer 24, the call is W096/04756 2 ~ 9 ~ ~ 7 PCT/GB95/01823 redirected out to a pre-programmed destination. ~hus, for each day, a customer may issue programming commands, such tha~, at predefined points during the day, calls are dir2cted to different numbers, thereby following a customer from location to location.

A switch 25 _s provided with 256 multiplex lines.

Thus, each of said lines includes two coaxial cables, for bi-~irectional communication, and each of said coaxial cabies is capable of conveving 32 channels, transmitted at 2 megabit per second. Thus, a plurality of coaxial communications are provided to the public switch telephone network, identified generally by reference 26 and other 2 meaaDit links are supplied to each of the respective per~pherals. Thus, in the example shown in Figure 2, 2 megabit links 27 are connected to peripheral 21, similar li,.xs 28 connect peripheral 22, while links 29 connect per pheral 23. Thus, each peripheral 21 to 23 etc. is arranged to receive voice channels in the form of 2 megabit multiplex connections. In addition, the pe- ~herals are arranged to receive control romm~n~ from the control computer 24.

Preferably, common channel signalling is used, such as signalling which conforms to the CCITT C7 recc~men~Ation and an interface for identifying and generating C7 signalling information is identified in Fig-lre 2 by reference 30.

The C7 interface 30, the central switch 25, the perlpheral devices 21, 22 and 23 communicate with the con~rol computer 24 via an ethernet connection, capable of conveying a data at a burst rate of up to ten megabits a second.

Calls are initiated in response to DTMF tones generated by a customer. In order to establish a call to the service node, signalling commands are transmitted over the common channei, by means of one of the channels forming part of those generally identified by reference 26, which, at switch 25, are directed to the signalling wos6/047s6 2 1 9 5 4 3 7 PCT/GB95/01823 - interface 30. The signalling interface 30 instructs the control computer 24, over the ethernet 20, to the effect that a calling customer reauires service. The signalling _nrormation will dentify to the control computer 24 an identification of the actual calling customer and, 'o some extend, an identification of the number called by said customer. Thus, on the basis of this information, the con~rol computer 24 instructs the switch 25 to connect the call to a specified channel connected to the required peripheral device. In addition, the control computer 24 aiso instructs the relevant peripheral as to the nature of the service required.

While a customer is making use of a service, it is common practice for the customer to generate signals in a form which is recognisable by the system. These signals may be interpreted by peripheral devices as part of the normal operation of the service. However, in addition, it mav also be necessary for customer generated signals to be recognised by the control computer 24, particularly if these relate to customer instructions which effectively terminate a service, such that the control computer is -equired to instruct switch 25 and the relevant peripheral to the effect that the service has now been terminated and, if re~uired, to allow the customer to make further use of the services available at the node.

In the telephony network illustrated in Figures l and 2, the service node includes a pluralit~ of peripheral devices in addition to switch 25 for connecting the peripheral devices to external commtlnicating channels 26.

~n order to establish calls and to monitor signalling while calls are in process, a significant level of processing capacity is required within the node. The service node is therefore provided with a control computer 24 which provides overall control of the system in an hierarchical manner. However, in addition to this, the switch 25 itself includes distributed processing devices which are themselves responsive to externally orisinating wo 96/047s6 2 t 9 ~- 4 3 7 PCT/GB95/01823 signals, sucA as DTMF signals generated by customers while calls are in process. Thus, the distributed processing devices are arranged to supply information derived from saic signals to the control computer 24. Thus, in this way, ,t is not necessary for the control computer 24 to scan incoming calls in order to detect customer generated signalling. The detection and identification of signalling C9m~n~ iS made locally at the switch 25 and data is only supplied to the control computer 24 over the ethernet 20, when a customer originating signal has been detected. Thus, the switcA 25 includes local processing devices responsive to external signals, such that information derived f~om said signals may be supplied to the control computer 24, in preference to the signals themselves. Thus, the computer 24 is relieved of the activity of scanning customer originated calls, which provides two significant benefits. Firstly, for a given switch size, less processing capacity is required on the par_ of the control computer 24. Alternatively, for a given processing capacity present on the control computer 24 a much larger switch may be used. The advantage of that is, for a given size of control computer 24, more central added value services may be included at the node, thereby effectively providing a greater economy of scale.

Secondly, given that it is not necessary for scanning to be effected by the control comput.~r 24, co ~l~.ication between the switch 25 and the control computer 24 is significantly reduced. Consequently, far less band width is -equired on the part of communication channels connecting control computer 24 and the switch 25 which, in this example, is effected by means of an ethernet. Thus, as can readily be appreciated, the level of band width available on the ethernet is significantly less than the level of band width employed to effect communication between the PSTN and the peripAeral devices, via the switch 25.

W096/04756 2 1 9 5 4 3 7 PCT/GB95/018~ Switch 25 is housed within an EMC l9" shielded rack, which stands 2 me~res high and contains a total of seven sAelves. One of these sheives is used to house power su~piy units for the ethernet .ransceivers, while the remaining six sheives are provided for component cards of the switch itself.

Within the switch 25 two types of shelf are provided -~A.ch are identified as line ~nterface shelves and swl~ching matrix shelves. In the preferred embo~i~ent, switching matrix shelves and interface line shelves are dupiicated so as to enhance reliability and a robust control computer, itself having duplicated shelves, prererably connected to the system, again enhancing overall reliability sucA that operation of the system is ~5 not dependent on a single card. Any of said cards may be removed wAile allowing the system to remain fully func~ional .

At the heart of the switch 25 is a non-~locking di_~_al switching matrix configured from commerc1ally avaiiable MITEL SMX/PAC chip sets. Switching matrix 32 receives 256 2 megabit per second coaxial pairs, cvnrorming to the ST-BUS recommenaation. Thus, each pair consists of a 32 channel multiplex for supplying information to the switching matrix 32 and a similar 32 channel multiplex for receiving information from said swi_ching matrix 32. The switching matrix 32 is arranged such that any channel on any of said multiplexed lines may be switched to any other channel on any other of said mul~ plexed lines. Generally, said communications are bidirectional, such that first of said pair is arranged to transmit data from location A to location B while the other pair provides co~ nication in the opposite direction, from location B to location A.

Furthermore, of tr.e ~hirty-two channel 2 Mbit/s multiplex, thirty of the channels are used to convey audio data, and of the remaining two channels, one is used to convey signalling data relating to the thirty audio channels, while the 3 7 W096t04756 PCT/GB95/018 remaining channel is used to provide frame synchronisation so that the position of each of the channels within the time mult plex may be determined by channel switching -i c-~itry Switching ma~rix 32 is controlled by an Intel 80286 mic-oprocessor based system identified generally by reference 33.

This microprocessor based system 33 is configured ~o contrcl _he switching matrix 32 in response to commands received over the ethernet line 20.

Signals transmitted through the switching matrix 33 conform to the ST-3US recommendation and may be considered as ~ volt TL compatible signals.

Signals transmitted over lines 26 to the PSTN and other lines connected to the peripheral devices, do not conform to this recommendation and may effectively be considered as analogue-like signals, given the reauirement for, in many situations, tr2nsmission over significant distances.

Thus, before signals may be supplied to the switching matrix 32 and before s_gnals emanating from said switching matrix may be supDiied ~o the outside network, it is necessary to per-orm a conversion procedure, which is effected via line interface c rcuits.

The rack contains four line interface shelves ident_fied in Figure 3 by reference 34.

Each line interface shelf mav have a total of ;6 line interface caras 37 and on each line interface card there is provided a resident Intel 80Z86 microprocessor system 38 and four bi-directional line interface circuits.

Thus, within the four line interface shelves there is capacity for 256 bidirectional interface circuits.

In this example, customer originating signalling erfected wnile calls are connected take the form of DTMF coaes. _ach DTMF code (dual tone multi-freauency) consists of two simultaneous tones, each selected from a different set of four possible tones.

Thus, there are 4 x 4 permutations of two tones, providing a total of 16 different codes which tne system may recognise.

However, W096/04756 2 1 ~ 5 ~ 3 7 PCT/GB95/01823 11 other modes of signalling may be employed consisting of audible signals of some form which may be detected by the svstem.

DTMF ~etection occurs on each line interface card under the control of the residenl Z86 computer system 38.

The nature of a particular DTMF detection will depend on the actual applicalion being made during a particular call connection.

The nature of DTMF codes required for particular applications are held by the control computer 24 and are accessible by systems operators.

In addition to specific DTMF codes, it is possible to program the control computer 24 to detect simply logical com~inations, such as X# followed by any of the numbers 1, 2 or 3 etc., as will ~e detailed subseauently.

When a detection seauence of this type is required for a particular connection, it is necessary for the detection commands to ~e down-ioaded to the respective line interface card.

This is achieved over the ethernet 20 and, initially, high leve! detection codes are supplied from the control computer 24 to the 286 microprocessor system 33 resident within .he switch shelf.

Switch shelf microprocessor svstem 33 _s responsible for identifying a particular line interface circuit and after receiving such information from the control computer 24 over the ethernet 20, the information is relayed to the respective 286 mic-oprocessor system 38 resident on the respective line interface card. mhus, the transfer of DTMF detection programming from 'he control computer 24 to a line interface card 37 occurs in two steps, firstly from the control computer 24 to the switch shelf microprocessor svstem 33 and secondly from the shelf cell processor 33 to the respective line interface card microprocessor system 38.

Thus, although it is possible for the control computer ~4 to communicate directly with the line interface sAelves, this is not done in practice as a hierarchical software structure exists where~y the control computer 24 communicates with the switch shelf W096/04756 2 1 9 ~ 4 } ~ PCT/GBg5/0l823 - microprocessor system 33, and the switch shelf microprocessor system 33 communicates with line interface card microprocessor system 38.

Thus as far as the control com~uter 24 is concerned all transfers to the switch 25 may be considerea as trar.sfers tO the switch shelf microprocessor system 33, thereby simplifying the way in whicA control computer 24 communicates with the switch 25.

A line interface card 37 is detailed in Figure 4.

The card includes 4 interface circuits 41, 42, 43 and 44, eacn connected to an input 2 megabit per second trunk line 45 and to an output 2 megabit per second trunk line 46.

The line interface circuits are responsible for converting between transmission compatible signals and 32 channel 2 megabit per second ST-BUS signals which are supplied to and received from a cross point switch 47, via 2 megabit per second input lines 48 and 2 megabit per second output lines 49.

The cross point switch 47 consists of a Mitel SMX/PAC switch, similar to switching chips employed within the main switching matrix 32. However, it should be appreciated that the c-oss point switch 47 does not effect switcAing on a call-~y-call basis. The cross point switch 47 is permanently configured during normal operation although said configuration may be modified to take account of upgrades and reconfigurations etc.

As shown in Figure 3, the line interface board 37 inciudes a 286 mic-oprocessor based system 38, connected to the ethernet link 20. The address and data bus of the microprocessor system 38 is connected to the configuration interface 39 of the cross point switch, such that said cross point switch 47 is configured in response to configuration data supplied from the microprocessor based system 38, as is known in the art. Thus, in response to this configuration, eacn ST-BUS output and input line (48, 49) is connected to similar lines 51 on the switch shelf side of the cross point switch 47. Thus, during normal operation, each channel supplied on an output line 48 from W096/04756 ~ ~q~ 3 7 PCT/GB95tO1823 13 the iine interface circuit from ~he cross point switch, wili be switched to a ~articular _hannel on a particular mu__iplex of the ST-BUS on the swltching matrix side of the cross point switch. Similarly, ~ parallel return path wi ' be provided througn the cross ~oint switch, thereDy allowing all calls to se directea ~hrough line in~erface ci--uits, ~ia cross ~oin~ switches resident on line in~erface cards, out to the main switching matrix 32 and then back to any selectea line _nterface card via a fur~her c~oss point switch based on said interface card ana out of .he main switcA 25 via a predetermined line in~erface circuit.

In addition to providing configuration codes for the cross point switch 47, _Ae 286 mic~oprocessor based system 38 is also capable of observing any communication channel being passed througA the cross point switch 47 by applying sui.able address codes to the aadress bus connected to interface ~0.

In addition to connecting -nlerface circuit side channels to switch shelf side channels, the cross point swi_ch 47 may also connect each nput line 48 rrom the line interface circuits 41 .o 44 ~o a respective aigital signal processor 52, 53, 54 and 55. Thus, the cross point switch 47 may be configured such that the input multiplex 48 rom line interface circuit '1 is connected to an output multiplex 56 supplied to a aisital signal processor 52, with similar connections being made to the other three sicnai processors.

Each digital signal processor 51 to 55 is a T~S320C31 device produced by Texas Instruments and is programmed to be able to detect DTMF codes on each of the 30 audio channels present on the 32 channel input multiplex 56.

It will be appreciated that signals conveyed through the cross point switcA 47, along with signals conveyed through the main switching matrix 32, are conveyed in digital form. The 2 mega~it per second ST-BUS is arranged to convey 32 time multiplexed channels. Thus, each W096/04756 2 1 ~ S 4 3 7 PCT/GBg5/0l823 channel Aas an iden~ifiaDle time slot within each frame and during this time slot 8 bits of data may be transmittea. Thus, when transmitting conventional speech signals, eacn of these 8 bit words convey a digital representation of an audio sample.

The aigital signal processors 52 to 55 receive information in the ST-BUS format and produce an output in similar format. However, the out?ut from each DSP does no~ represent an audio signal but represents an indication of information contained within an audio signal; in particular, the presence and identification of DTMF tones.

Thus, in the previously suggested configuration, DSP 52 examines 30 speech channels received via line interface circuit 41, while said channels are being supplied to the switching matrix 32 via the switch side communication buses 51. The digital signal processor 52 is programmed to effect.vely convert signals received in the time domain into freauency domain signals, and then to provide an inaication when these frequency domain signals correspond to .ecognised DTMF tones. Thus, the digital signal processors produce outputs indicating the presence of DTMF codes.

A DTMF signai may take one of sixteen states, therefore four output bits are produced by a DSP for eacn audio channel in order to identify a particular DTMF detected code.

In aaditlon to this, a further bit is used to provide a vaiid code detected signal and the microprocessor 38 is arranged to ~A~ine these five bits of data provided by each DSP on each of the 30 channels examined by each DSP .

Thus, during communication, a customer using a channel suppiied to line interface circuit 41 may depress a key resulting in a DTMF signal being supplied over this channel. The audio signal may convey audio information in addition to the DTMF code, but by implementing a fourier transform on each of the thirty audio channels supplied to each of the DSPs, it is possible to detect the presence of W096/04756 2 1 9 ~ 4 3 7 PCT/GB95/018~ the recognisable DTMF tones and successfully return this information to the microprocessor system 38. Thus, when the DTMF tone code is generated, the DSP 52 detects the presence of ~his DTMF tone and returns the rele~ant four bil representation to the cross ?oint switch 47 on a respecti~e channei associated with the particular channel being supplied via the line interface circuit 41. The DSP generates the five bit DTMF recognition signal throughout the time ~hat the DTMF tone is being recognised.

10 - Generally a DTMF tone must be present over a period of time before it is considered valid by the system, so that spurious noise signals or speech cannot wrongly trigger recognition of a DTMF code.

Thus the DSP 52 will pause after recognising a DTMF tone combination, for long enough to ensure that it has been deliberately intended.

Thereafter the DSP 52 generates the five bit recognition code once for each corresponaing audio sample, and supplies it to the cross point switch 47.

Thus, when a tone is _orrectly recognised by a DSP, there will be a period of time during which the recognition signal is continuously generated.

Within .his time period the mic-oprocessor 38 must interrogate the cross point swltch to capture DTMF recognition signais as they occur in real time.

A successful recognition of a DTMF tone by a DSP will result in the five bit signal being generated for a minimum period of 32 milliseconas.

Thus, in order to ensure successful capture of all -ecognised DTMF codes, the microprocessor must sample a DSP output channel once every 16 milliseconds.

In order to successfully capture DTMF recognition for any of the thirty channels connected to a DSP, all thirty must be sampled within the 16 millisecond time frame.

Thus, in crder to ensure reliable capture OI all DTMF recognitions within the line interface card, the microprocessor 38 must sample all 120 audio channels within each 16 miliisecond time frame; and this W096/04756 ~ 1 9 5 4 } ~ PCT/GB95/01823 16 represents a substantial computational requirement of the microprocessor 38.

In the prefer-ea embodiment, each sample of an output channeî ~om a ~SP is placed into a circular buffer _ according .o the rules defined in Ihe flowchart shown in Figure 5. ~hus, -lthough the intensive task of sampling all the audio channels is performed at the required rate, the processing of the resulting data may be performed over a longer .ime 2eriod by more complex software. ~he flowchart in Figure S details the operations of an interrupt ~outine running on the microprocessor 38 which is callea at sixteen millisecond intervals.

Thus, regardless of whatever functions the microprocessor is otherwise doing, .his routine will be called every sixteen milliseconds.

Step l5l signifies the generation of the sixteen millisecond periodic interrupt, which will force the microprocessor 38 to suspend other operations and execute the nterrupt rou~ine. ~t step 152 the channel index N is reset to zero.

The channel index will be used subse~uently to define access .o a particular channel within the cross point switch 47 wnich is used .o convey data from a channel output from a DSP.

The channel index is also used to access the circular buffer and other memory locations within the 286 microprocessor system 38 which are associated with a particular audio -hannel.

At step 153 bit 5, the validation bit, of channel(N) is co~n~red with its previously recorded value to see if - it has changed. f bit five is unchanged, the other four bits of the DTMF recognition code are either valid or invalid.

Either way, no actions needs to be taken for the current cnannel.

However, if bit five has changed, step 154 decides whether bit five indicates a valid or an invalid signal.

If bit five indicates an invalid signal, a valid recognised DTMF tone has just ended and step 155 may place the previously valid four bit DTMF recognition code into a circular buffer for the current chAnnel.

In W096/04756 2 1 9 5 4 3 7 PCT/GB95/0l823 other woras, valid DTMF tones are only recorded by the microprocessor 38 at the end of their duration. A circular buffer is one where a number of memory locations, for example 64, is indexed by a counter which automatically increments every ~ime a data item is written into the location wnich is indexed by it.

The counter has the property that only a pre-defined number of bits are used for the index, so ~hat incrementing the counter beyond 63 is impossible.

Thus, computational overhead and memory are 10saved.

The disadvantage is that another counter must be usea to read out values from data locations within the buffer before it is overtaken by the buffer input counter.

suitable size for the circular buffer may be defined, such that, siven a set of reasonable operating conditions, 15it is exzremely unlikely for this to happen.

In the event that step 154 decides that bit five indicates the presence of a valid tone, no significant action need be taken.

Step '~6 inc-ements the channel index N, so that the 20steps 153, 154 and 155 may operate similarly on data associatea with _he next channel.

Before doing so, however, step 157 ensures that 'here are still audio cAannels to be processed.

If there are some left (N is less .han 120) control is returned to step 153.

If all 120 25audio channels have been sampled, control is returned to the process which was being executed at the time the interrupt was called.

The interrupt routine shown in Figure 5 represents a highly erficient way of sampling, interrogating and 30storing DTMF recognition codes ,rom a large nl~mher of channels in real time using a microprocessor 38.

Once DTMF codes have been stored in circular buffers, the microprocessor 38 may analyze the stored codes tO see if the pattern of codes meets conditions defined by a scan 35pattern.

Thus the microprocessor 38 on the line interface card is able to record and match DTMF patterns against a scan pattern provided to it by the microprocessor 33 in W096/04756 2 t ~ ~ 4 3' ~ PCT/GB95/01823 the switch shelf, ~nich has received the scan pattern and associated control nstructions from the control computer 24.

Furthermore, Ae scan pattern may specify a range of possible condit_~ns, so that the microprocessor 38 may recognise a varie~v of incoming DTMF code patterns as meeting the reaul-ed conditions, rather than simply searching for an -dentical match between a pattern and incoming DTMF codes. Thus a limited form of pattern recognition may be ~erformed by the line interface card processor 38.

The procedure shown in Figure 6 outlines the operations performed in different parts of the service node in order tc set up and respond to a DTMF scan pattern.

At step 161 the control computer generates a lS command stream for a particular customer call.

The comm~n~ stream includes .he scan pattern, which is an embedded string of cAaracters. These characters define the set of conaitions which an incoming DTMF sequence from the customer must satisfy before said sequence is considered to match a partic-iar pattern.

The format of the scan pattern is descr bea later.

The comm~n~ stream is passed over the ethernet interface to the switch shelf microprocessor 33. At step 162 the switch shelf microprocessor 33 redirects the command stream, again over the ethernet, to a line interface card.

At step 163 the line interface card microprocessor 38 interprets the co~mAn~, deriving signalIing information sucA as the channel number from the co~n.~ stre_m.

Also at step 1_3 the microprocessor 38 translates the scan string into a different format in order to provide maximum speed efficiency during the process of attempting to match a DTMF ccae sequence, which occurs at step 164.

At step 164, the DTMF codes stored in the circular buffer associated with the customers channel are csmpAred with the speed-efficient scan pattern to determine whether the conditions are met.

If these conditions are not met, step 164 loops back to its beginning again.

W096/04756 ~ ~ 9 5 4 3 7 PCT/GB95/0l823 When a DTMF sequence in the circular buffer matches a scan pattern, con~rol is diverted to step 165, where the line card microprocessor 38 generates a response stream, inciuding the matching DTMF sequence.

The response stream Sis sent over the ethernet 20 to the switch shelf microprocessor 33, wAich at step 166 directs the response stream to the con~r~l computer 24 over the ethernet.

At step 167 the ccntrol computer uses the information contained in the response stream to direct further actions 10that are required lo complete the customer service.

The control computer 24, having sent a command stream to the switch sheif, may return to other tasks until the remote processors send back a positive response stream.

Thus, sophisticatea DTMF pattern matching is provided 15without the danaer of computational resources being overwhelmed by a coincidence of numbers of callers requiring a similar service at the same time.

As described previously, the control computer sends a cs~m~n~ stream containing a scan pattern defining the 20conditions for a ~ositive match.

This scan pattern is a string of characters in a relatively legible format to aid service software ~riting and debugging.

The format includes specifiers, and is defined in the following way:

A constant specifier may be used to define specific 25matching characters. A constant specifier may be 0 through - 9, A through D, - or ~.

An example of this is "X#22", which reauires z _ustomer generated DTMF sequence of "X#22" to achieve a positive match - it must be identical.

An e~ual spec fier defines that the DTMF incoming 30character must be the same as the one which preceded it.

For example "X=2=" ~equires "X#22".

A variable specifier follows the form %{n.{m}}p.

Where p is one of "~", "n", "m" or "[digits]".

"a" specifies any DTMF digit.

35"n" specifies any numeric DTMF digit 0-9.

"m" specifies any numeric digits terminated by #.

W096/04756 ~ 3 7 PCTIGB951018 "[digits]" specifies any DTMF digit in the string "digits", except wAen the first character is a ~, in which case it specifies any DTMF digit except those in the rest of the string.

The optionai ~aiue n (as in %{n.{m}}p.) can be used to specify detection of a fixed number of digits. ~f only n is specified as the template string, a match is always made on entry of _he nth DTMF digit by the caller.

The optional value m (as in %{n.{m}}p.) can be used to specify a maximum number ~, DTMF digits to be input by the customer.

For example "%5m" is matched if an input of 5 numeric digits terminated bv a "~" are detected.

"%3.5m" is matched if an input of 3 to 5 numeric DTMF digits terminated bv a "~" are detected.

"%1[123]" is matched if a single DTMF digit "1", "2" or "3" is detected.

"X=%1[123]=" is matched by the patterns "##11", "~#22" or "X#33".

In aadition r3 the above methods of specification, a logical OR character (,) may be used to specify a plurality of matching conditions, such that only one of the conditions contained within the string need be met~lo achieve a positive match.

This character is represented on some computers as a single horizontal line (without the usual embellishments associated witn letter I or numeral 1), shown herein as "'", and has the ASCII value 124.

~hus using the logical OR character a template definition may be constructed thus:

"#=%1[123]='~ 12345],%10.20m'%1t~0]" which enables the customer to generate one of four different possible tYpes of response.

Whichever response the customer generates will be passed back to the control computer ~n the response stream, and thus the control computer 24 may initiate one of four different possible responses to the customer's DTMF sequence.

In this way an intelligent dialogue may be built up with the caller, so W096/04756 2 ~ S 5 ~ 3 7 PCT/GB951018~ 21 that the caller may navigate through a menu driven service.

In the commAn~ string generated by the control computer, variables and flags are provided to define various aspects ~f the process OI DTMF pattern recognition.

Two variables, known as "first_dig_to" and "inter_dig_to", define the maximum time before a first digit is received and the maximum time between subsequent - digits respecitiveiy.

When inter_dig_to is exceeded, the DTMF codes may ~e successfully matched against scan patterns defining a "X" as the last digit, even if this has not been provided by the caller.

When first_dig to is exceeded, the operation is defined by the CONTINUOUS_SCAN flag defined below. The flags included in the comm~n~ string are:

CLEAR_~ Causes any previously detected but unchecked DTMF disits to be discarded on initiation of additional pattern recognitions.

FIRST_DIGIT NOTIFY - Causes the line card interface microprocessor 38 to generate a response stream on receipt of the first DTMF cigit code after a pattern recognition has been initiated.

CONTINUOUS SCAN - Causes the operation to restart automatically either after a successful pattern match, or after the time out aefined by the "first_dig_to" has been exceeded.

Thus a commana stream for initiating.a DTMF pattern recognition incluaes a template specification string, variables and flags, and associated channel, and service identification data. A response stream from the line interface card mic.oprocessor 38 assigned the task will include the actual 3TMF codes matching the scan string string, and associated channel and service identification data, so that the control processor may associate the response stream with the original command string and a particular service or a particular connected caller.