ABSTRACT
We present the design of a
microcontroller-based system for intelligent telephony. In addition to providing the ability to play
different messages in response to distinct calling parties, our system combines
calling number identification with digital recording and playback capabilities
to permit audible announcement of the calling party between rings, using the
calling party's voice. We discuss
implementation issues and plans for future prototypes using newer
technology.
1.0
Introduction
With
the introduction of Combined Local Access Signaling Service technology, or
CLASS, telephone companies are providing an increasing number of powerful
services to subscribers [1]. One such
service, Calling Number ID, permits the identification of the number of the
calling party at the receiver [2]. We
have designed a system for intelligent telephony that makes extensive use of
this technology, to provide features not found in many existing devices.
We
present an overview of system functionality, discussing basic subcomponents and
their interaction. Software and hardware
are described, along with suggestions for future work.
2.0
System Functionality
Our
device is capable of a variety of functions, including
1) Calling number identification. Our system is equipped to identify and display
the number of the calling party if the phone network supports calling number
ID.
2) Caller-specific answering messages.
If calling number ID is supported by the network, the system can play an
answering message specific to the number of the calling party.
3) Call database management.
The system maintains a database of recorded calls, including the number,
time of call, and message left. Any
information in the database may be accessed through the keyboard. Database size is limited by available memory;
the current prototype can store information for 100 calls. The device also contains an autodialer, and
can dial any number in the databook with the push of a button.
4) Caller-specific audio identification. The device can play a prerecorded message on a
loudspeaker, based on the number of the calling party. Thus the voice of the calling party can be
heard in between rings, announcing the identity of the caller.
3.0
The Prototype System
A
block diagram of the device is shown in Figure 1. It consists of three subsystems: a telephone
interface, a speech subsystem, and a microcontroller/memory subsystem. Each of these is connected to an 8-bit data
bus. An analog output is used by the
speech subsystem and telephone interface.
The
device itself connects to a standard telephone network and telephone. Commands are input through a keyboard, with
characters output on a liquid crystal display. A microphone and speaker are
also provided for recording and playback of messages, along with an autodialer
for the rapid dialing of databook numbers.
3.1 The Telephone Interface
The
telephone interface connects the telephone and network with the rest of the
system. A diagram of the telephone
interface is shown in Figure 2. The
telephone and network connect to an FCC Part 68 approved Data Access
Arrangement circuit, or DAA. The DAA in
turn has a bidirectional connection to the system audio output. The network also connects to a caller ID
circuit, or CID. When calling party
information is detected on an incoming call, the resulting bit-serial stream is
placed on the data bus for consumption by the microcontroller. Finally, the telephone interface includes the
dialer, which has a unidirectional connection to the system audio output.
3.2 The Speech Subsystem
The
speech subsystem is responsible for all recording and playback of speech
signals, as shown in Figure 3. An 8-bit
PCM CODEC provides A/D and D/A signal conversion. Digital CODEC values can be written to or
driven by a 1-Mbyte RAM, addressed by a 20-bit counter. Portions of this counter can be loaded from
the data bus if the appropriate driver is enabled, to select appropriate
messages for recording and playback.
The
RAM stores all system messages, both incoming and outgoing. The number of messages and the length of each
are a function of available memory.
Based on an 8KHz sampling rate for 8-bit samples, we obtain _130 seconds
of time for each megabyte of memory.
Tradeoffs may be made for with regard to the number of messages and
message length at the convenience of the user.
3.3 Microcontroller/Memory Subsystem
Like
other systems for intelligent telephony (see for example [3]) our device is
microcontroller-based. The
microcontroller and its attached memory are shown in Figure 4.
The
microcontroller executes instructions from the ROM chips, while the RAM holds
the database and other runtime-dependent values. Currently, both RAM and ROM
chips are 8Kx8. The microcontroller
issues 16-bit addresses, the upper 3 of which are decoded to activate various
control points in the machine: the RAM and ROM chip select lines, the keyboard,
the LCD display, the message RAM counter, and a parallel I/O controller (not
shown). The remaining 13 bits are sent
directly to the ROM and RAM chips.
4.0
System Operation
The
microcontroller executes instructions stored in ROM. All system code fits in one 8Kx8 ROM; the
other is currently unused The system
operates in two modes: automatic mode and user mode.
4.1
Automatic Mode
When
the machine is not being employed by the user, it monitors the phone line for
an incoming call. The device handles
these calls in the following manner:
1) Ring signal detection. The
DAA is polled for the occurrence of a ring signal. Once one is detected, the CID circuit is
powered up and activated.
2) Calling
number identification. The CID
circuit extracts the caller-ID string from the phone line and places it
bit-serially on the data bus. The
microcontroller processes this information and looks for entries in the
database that match it. If the
information is garbled or unavailable then the database is not searched. This could be caused by noisy phone lines,
connection to a non-CLASS network, or the blocking of ID information by the
calling party [4].
3) Caller-specific
audio identification. If a matching database entry is found, the
microcontroller retrieves its address in the message RAM and loads the message
RAM counter. The announcement message
associated with the calling number is then played back from RAM through the
CODEC to the speaker, in between each ring.
4) Caller-specific
response. After a user-selectable
number of rings, the phone is forced off hook by the microcontroller, and the
answering message stored in the appropriate databook entry is played back..This
time, the output of the CODEC is sent through the DAA to the phone network.
5) Recording of message. Any message from the caller is recorded
in the message RAM. The data and time of
the call are stored in the database.
At
any time during this process, the user may pick up the phone. This cancels all remaining operations.
4.2 User Mode
The
system may also be interrogated by the user through the keyboard. The user can examine the database, add or
delete entries, dial entries with the push of a button, and play back all
messages associated with the database. Autoanswering capabilities can be
disabled, the time can be set, and all messages associated with incoming calls
since the last use of the machine can be examined.
5.0 Hardware Description and Future Work
Our
device measures 14" x 12" x 3", about the size of a conventional
answering machine.The design was developed using the techniques described in
[5]. The prototype is a wire-wrapped
design containing approximately 40 discrete components. We estimate the size could be reduced by a
factor of two by replacing logic chips with a single field programmable gate
array, a special kind of user-programmable integrated circuit that can replace
large numbers of chips [6]. Further
reductions might be possible through the use of application-specific integrated
circuits, and by replacing the microcontroller with a more advanced
processor. A truly aggressive approach
employing full custom VLSI design might produce a prototype the size of a large calculator.
Future
improvements include backup battery circuitry so that database contents are not
lost on power down, and the addition of more memory to increase database size
and message length.
6.0
Acknowledgement
The
authors would like to thank Doug Dapice and New England Telephone for their assistance
in procuring a CLASS network simulator.
We also thank Professor Yves Thomas and IRESTE at the University of
Nantes for their support of the Dartmouth/Thayer exchange program. Finally, we
thank Thayer School Staff Engineer Doug Fraser and Tad Truex for their
technical expertise and assistance.
This
project was designed at the Thayer Rapid Prototyping Facility, a laboratory for
the rapid design and construction of electronic systems.
7.0
References
[1] Perry, T.S., "Telephone
Challenges: A Plethora of Services",
IEEE Spectrum, July 1990, pp 25-28.
[2] CLASSSM Feature: Calling Number Delivery, Bellcore
Technical Reference TR-TSY-000031,
Issue 3, January 1990.
[3] Ooi, T.H. and Lau, K.T.,
"Enhanced Telephone Set on the Inter-IC Bus", IEEE
Transactions on Consumer Electronics, Vol.
36 No. 1, February 1990, pp 18-22.
[4] CLASSSM Feature: Calling Number Delivery
Blocking, Bellcore Technical Reference TR-TSY-000391, Issue 2, January 1988.
[5] "Embedded Real-Time
Systems: A Specification And Design Methodology", John Wiley Publishing, October 1992.
[6] Fagin, B., "Quantitative
Evidence of FPGA Utility in Special and General Purpose Processors", Journal
of VLSI Signal Processing Special
Issue on Field Programmable Gate Arrays, 1993, in press.
