 |
Description  |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to a method for the recognition of a command supplied
to an interactive system by a user of the system, in which in one section
of a program
(a) the system asks the user of the system a question in the form of a
speech signal,
(b) next, a command recognition unit is switched on for recognition of a
command to be supplied to the system by the user,
(c) depending on the result of the recognition procedure in step (b), and
on the basis of a decision model, the system makes a choice of one
reaction from a group of at least two possible reactions to the command,
and
(d) next, this reaction is then presented to the user by the system.
2. Description of the Related Art
The invention also relates to an interactive system for the implementation
of the method.
An example of the method of the type mentioned in the preamble is shown in
FIG. 1. The method could, for example, be implemented in an automatic
ticket issuing machine for obtaining train or bus tickets, with the mutual
transfer of information taking place by means of the spoken word. The
method begins at block 10 and runs to block 11 in which step (a) is
carried out, that is to say, the system asks the user/buyer the first
question, for example:
question 1: "Do you want to buy a train ticket or a bus ticket?" In
addition, step (b) is then carried out: the command recognition unit is
switched on. After that, the user gives his first command in block 12 (or
in other words: his first answer), for example:
command 1: "(I want) a train ticket".
The decision model is indicated with block 13. The decision model chosen is
very simple. It can only select from two reactions. One reaction is that
the program runs back via the circuit 14 and the block 15 to block 11.
This path is chosen if the command recognition unit has not recognized the
command which was given. In block 15 the system then gives, for example,
the following information:
"Your command was not recognized. I shall repeat the question."
If a command is in fact recognized (and let us assume for a moment that the
command has also been recognized correctly, namely the command that the
user has also given--which need not necessarily be the case!) then the
program runs to block 16, in which the next question is asked by the
system, for example:
question 2: "What is your destination?" To which the user could, for
example, answer in block 17:
command 2: "Amsterdam".
Next the decision procedure in block 13' again takes place, which can lead
to the program running via the circuit 14' and the block 15'--in which the
same information as in block 15 is given to the user, namely that the
command has not been recognized--to block 16 for the repetition of the
question. When a command is recognized the program runs to block 18 for
the third question:
question 3: "What station do you want to go to?" Which in block 19 could be
followed by:
command 3: "Amstel station". The fourth question might be:
question 4: "Do you want to travel first or second class?" To which the
answer in block 21 is, for example:
command 4: "Second class".
The blocks 13" and 13'" again indicate the decision phases and 14" or 14'"
and 15" or 15'", respectively, indicate the circuits that are gone through
if no command is recognized.
In block 22 the system indicates the price of the ticket.
In block 23 the buyer pays, so that in block 24 he receives his ticket,
after which the program stops in block 25.
In FIG. 1 the program sections 26, 26', 26", . . . etc. are indicated with
the dashed lines. The program sections 26 and 26' have block 16 in common.
The program sections 26' and 26" have block 18 in common, . . . etc.
It should be mentioned here that program sections may also contain larger
units than the units 26, 26'. . . .
The method shown in FIG. 1 has a great advantage. If the recognition is
correct in all sections of the program, then the user very quickly obtains
a ticket. This is because the system asks a following question immediately
after recognizing a command and does not first give further information to
the user relating to the command which it has recognized.
The method shown in FIG. 1 also has a number of disadvantages. For
instance, it may occasionally happen that the user discovers only when he
or she arrives at program section 26'" that in an earlier section of the
program the system has recognized a different command from the one he or
she gave, so that the user is now in a completely different branch of the
dialogue. For example, he/she wanted to buy a bus ticket and only found
out in program section 26'" that he/she was in the dialogue for obtaining
a train ticket.
Another disadvantage is that because of its simplicity the system is fairly
rigid. A command is either accepted or it is not. In the latter case the
question is always repeated, which can be very annoying.
Yet another disadvantage lies in the fact that in changing circumstances
during the command recognition, particularly for example in those cases
where the conversation between a user and the system takes place via the
public telephone network--the transfer (path) from the mouth of the
speaker/user to the input of the system changes in terms of time--the
system cannot adjust its reactions to these changing circumstances.
SUMAMRY OF THE INVENTION
The purpose of the invention is now to provide a method or a system that
functions better, is better adapted to the user and can adapt to changing
circumstances, while maintaining the advantages of the known methods or
systems. For this purpose the method according to the invention is
characterized in that the user presents protests to the system, as a
result of the fact that a command has been recognized wrongly by the
system, add the number of protests is counted for the relevant program
section and possibly earlier program sections. If the number of protests
exceeds a certain threshold value the decision model is adjusted at least
during a certain time interval in such a way that in step (c), if a
command is recognized, in at least a number of the cases the decision
model chooses a reaction in the form of a more explicit indication for the
user of the command that has been recognized by the command recognition
unit.
The invention is based on the realization that the system should be
adaptive and in its reactions to the user should be able to adapt itself,
on the one hand to the user and, on the other hand, to changing
circumstances (for example in the transmission path between user and
system).
The adaptation to the user can, for example, be necessary because the
latter speaks fairly unclearly, while the system nevertheless recognizes a
command--and hence a wrong command. In most cases, after the system has
asked the next question, the user will start to protest. If this happens
frequently the system must, for example, build in an extra check in the
dialogue. Also in the event of the transmission (path) between the user
and the system being poor, the system will often make a mistake in
recognition which likewise leads to protests on the part of the user. Now,
by counting the number of protests and, if the number of protests exceeds
the above-mentioned value, by adjusting the decision model in such a way
that a during a certain time interval a more explicit indication of the
command which has been recognized is given to the user, the user acquires
a better possibility of checking with regard to the command recognized by
the system.
The user then notices more quickly if something goes wrong or threatens to
go wrong. On the other hand, if it turns out after the above-mentioned
time interval that the circumstances with regard to recognition have
improved, it is possible to switch over again to a less explicit
indication, for example no further indication, as in the method shown in
FIG. 1, so that the program again runs faster there and consequently the
user does not become irritated (or does so less quickly).
The above-mentioned time interval may, for example, be a fixed time
interval. For instance, for two minutes after the number of protests has
exceeded the threshold value the system will give a more explicit
indication of the command which it has recognized. After the two minutes
the system then returns to the original decision model. It is also
possible for the system to return to the original decision model when,
after subsequent questions from the system, for example, correct
recognition has taken place twice without any protest from the user. In
this case the time interval is therefore related to the time within which
the system and the user have dealt with the next (two) questions. The time
interval here is therefore not fixed.
The said value which must be exceeded in order to change the decision model
can be chosen at random. This value will, of course, be larger than or
equal to two. If the value is greater than two the aim will again be not
to choose it too large in order to prevent it taking too long (too many
protests being needed) to change the decision model. A suitable value may
lie between two and, for example, five. Yet another possibility is for the
more explicit indication to be given by the system only within one program
section (or one or more consecutive program sections). The time interval
then depends on the duration of the program section(s).
In accordance with the method according to the invention, starting from the
decision model and depending on whether the protest score of a command
recognized by the command recognition unit exceeds a certain threshold
value or not, the system can make a choice between a first and a second
reaction in step (c). In the first reaction the system asks the user a
following question immediately after recognition of the previous command.
In the second reaction the system gives the user initial information about
the command which the command recognition unit has recognized. The method
is characterized in that when the number of user protests reaches a value
at which it exceeds the said threshold value, the threshold value is
increased at least during the aforesaid time interval.
In favorable recognition circumstances, at a score that lies just above the
threshold value the first reaction will be continued, a following question
being given immediately after recognition, until the score reaches a
raised threshold value at which the second reaction will be provided.
However, in deteriorating circumstances, when the score reaches the first
threshold value the system will immediately provide the second reaction in
which it gives initial information about the command which was recognized
by the system. In this way an extra check is built in, which is necessary
in the case of deteriorating circumstances.
It should, however, be mentioned here that for commands which have been
recognized with a score below the original threshold value, and for
commands recognized with a score above the raised threshold value there is
no change in the reaction of the system in deteriorating circumstances. It
is only in a number of cases that the reaction of the system changes
depending on whether there are deteriorating circumstances, namely in
those cases where the score of the command which has been recognized lies
between the original threshold value and the raised threshold value. It is
therefore only in those cases that a more explicit indication of the
command which has been recognized is supplied to the user only if there
are deteriorating circumstances.
The method can also be characterized in that if the number of protests
exceeds the said threshold value, then during the time interval the system
always reacts by giving the user information about the command which the
command recognition unit has recognized. This method can be applied, for
example, in a system as described above in which, before the number of
protests has exceeded the said value and depending on the question of
whether the score exceeds a certain threshold value or not, the system can
make a choice from the first reaction (after recognition the next question
is given immediately) or the second reaction (after recognition, initial
information is given about the recognized command). After the number of
protests has exceeded the said value the system then switches over to only
the second reaction as a reaction to a command which has been recognized.
Thus, after recognition of a command, there is a change in the reaction of
the system only for those cases for which the protest score of the command
which has been recognized lies above the above-mentioned threshold value.
The method can also be applied in a system in which, before the number of
protests has exceeded the said value, the reaction of the system involves
giving the next question after recognition of a command. After the number
of protests has exceeded the said value, the system, after recognition of
a command, always gives the information about the command which has been
recognized.
In both the latter two cases the extra check is therefore not obtained by a
shifting threshold, which means that in the case of commands for which the
scores are above the shifted threshold a still further check cannot, or
will not, take place. In the previously described cases an (extra) check
exists for all commands during the time interval. If a command has been
wrongly recognized then the user protests, as a result of which the
question can then be repeated.
Another possibility is that the system, in reaction to the user, asks the
question as to whether the command recognized by the command recognition
unit is also the command given by the user. In this case, the command
recognition unit can then be further equipped for the recognition of a
following affirmative or negative answer from the user. In response to a
negative answer from the user the system can then repeat the original
question.
A further refinement of the method can be obtained by extending the time
interval if, within such time interval, a number of protests are again
counted such that the said threshold value is exceeded. The extension can
be made in such a way that after the said value is exceeded within the
current time interval, the time interval starts running again from the
moment at which the threshold value again was exceeded.
An interactive system for carrying out the method according to the
invention, is equipped with
an activation unit for asking the user of the system a question in the form
of speech,
a command recognition unit for the recognition of a command supplied to the
system by the user and containing a decision unit, and
a command status unit in the activation unit coupled to the command
recognition unit for delivering to the user one reaction from a group of
at least two possible reactions, depending on the decision model in the
decision unit. Such system is characterized in that it is further equipped
with:
a protest recognition unit for the recognition of a protest supplied to the
system by the user if a command is wrongly recognized,
a counter coupled to the protest recognition unit for counting the number
of protests supplied during a relevant section of the method program and,
if applicable, previous sections of the program, and
a comparison circuit coupled to the counter for comparing the number of
protests with a threshold value, and for producing a control signal if the
number of protests exceeds this value. The control signal is supplied to
the decision unit, which is equipped for adjusting the decision model, at
least during the above-mentioned time interval after receiving the control
signal, so that it causes the actuation unit to give the user a reaction
in the form of a more explicit indication of the command which has been
recognized by the command recognition unit.
The system can be further characterized in that the decision unit is also
equipped for increasing the threshold value, at least during the time
interval, after receiving the control signal. As a result, the threshold
value is raised after the number of protests has exceeded the initial
threshold value.
Another possibility is that the system is characterized in that the
decision unit is equipped in such a way that at least during the time
interval after the reception of the control signal the activation unit
reacts by giving the user information about the command which the command
recognition unit has recognized and in that furthermore the command
recognition unit is also equipped for the recognition of an affirmative or
negative answer from the user.
The system can be further characterized in that the decision unit contains
a detection unit for the detection of a control signal that occurs within
the time interval after the occurrence of the previous control signal, and
in that the decision unit is equipped for extending the said interval
after detecting the second control signal. In this way, the time interval
can be extended if the said threshold value has again been exceeded.
The system, in which for the recognition of a command supplied from a group
of possible commands, the command recognition unit compares the command
supplied with the reference pattern of a command from the group of
commands, which reference patterns are stored in a memory in the command
recognition unit, can be further characterized in that, during the
comparison, the sequence of the commands of which the reference patterns
are compared with the command supplied corresponds to the sequence of the
probabilities of the occurrence of the commands, in such a way that the
reference pattern of the command with the highest possibility of
occurrence is first compared with the command supplied. An acceleration of
the recognition procedure is achieved in this way.
BRIEF DESCRIPTION OF THE DRAWING
The invention will be described with reference to the drawings. These show
the following:
FIG. 1, the known method
FIG. 2, a first example of an embodiment of the method in accordance with
the invention,
FIG. 2A, a section of the embodiment example of FIG. 2, but slightly
modified,
FIG. 3, a second example of an embodiment of the method,
FIG. 4, two different decision models (a) and (b) on the basis of which the
command recognition unit makes a choice of a reaction from a group of at
least two reactions depending on the user protest score during the command
recognition,
FIG. 5, an example of an embodiment of an interactive system for the
implementation of the method of FIG. 2,
FIG. 5A, the system of FIG. 5, but slightly modified, for the
implementation of the method of FIG. 2A,
FIG. 6, an example of an embodiment of an interactive system for the
implementation of the method of FIG. 3, and
FIG. 6A, another example of an embodiment of an interactive system.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
FIG. 2 shows a first embodiment of the method in accordance with the
invention. FIG. 2 shows a section of the method program, such as, for
example, the program section 26 in FIG. 1, but further modified with the
characteristic components such as are necessary in accordance with the
invention. The program runs from the line 30 to the block 31 where the nth
question is asked, for example:
"Do you want to go to Amsterdam, Rotterdam or Utrecht?" The responsive nth
command is given by the user in block 32, for example:
"Amsterdam".
The decision model 33 is such that, just as in FIG. 1, the program either
goes back to block 31 via line 34 and block 35, because no command
whatsoever can be recognized by the command recognition unit--in block 35
the system then gives the same information as the system of FIG. 1 in
block 15--or after recognizing a command goes on to block 36 where the
following n+1th question is asked, for example:
"Amsterdam Central Station or Amstel Station?"
If the system asks the above question at the n+1th question, while the user
had given "Rotterdam" instead of the command "Amsterdam" as the command in
block 32, then the recognition unit has obviously performed a faulty
recognition. The user will realize this at the moment when the system asks
the n+1th question since the above-mentioned stations simply are not
located in Rotterdam.
At this moment the user will start protesting, block 37. The system is
equipped to recognize these protests, for example because the system has a
special "protest" button which the user presses at this moment, or because
the system is equipped for the recognition of an acoustic protest--for
example, the speaker utters the word "wrong". If no protest is recognized
the program runs on via block 38 in which the n+1th command is given. If
there is a protest, however, then the program runs on via the branch 39 to
block 40. Here a counter is raised by the value of one. Next, the system
announces in block 41:
"Your command was apparently not correctly recognized. I shall repeat the
second-last question."
The program then runs via branch 42 to block 43 in which it checks whether
the content m of the counter exceeds a certain value v. If not, the
program then runs via branch 44 to block 31 for the repetition of the
second-last question.
Now it may be that in one or more (preceding) program sections the user has
uttered a protest so many times that it turns out in block 43 that m has
become greater than v. In that case the program runs via block 46--in
which the counter is reset to zero--to block 47--in which the decision
model is adjusted in block 33--and 48--in which the system announces:
"Because of the poor recognition results, after each command you will be
given an indication of which command has been recognized"--to block 31.
The decision model 33, and also the decision models (such as 33') in
successive program sections, is or are now changed for at least a certain
time interval, in the sense that if a command is recognized the program
runs on via circuit 49 to block 50. In this block the system accounces:
"The command--Amsterdam--has been recognized" or the command which has been
recognized: "Amstersdam" is itself repeated.
The user now has an earlier opportunity--namely, even before the n+1th
question is asked--to protest, see block 51. If no protest follows, the
program runs on to block 36. If a protest does follow, the program runs
via branch 52 to block 40--in which the counter status is again raised by
one, and block 53 in which the system can announce:
"Your command was apparently not correctly recognized. I shall repeat the
question."
Next the program runs back via block 43 to block 31. If the time interval
expires and the counter status has not again exceeded the value v in that
time, the decision model then switches back to the old situation in which
the program continues via branch 57 after the recognition of a command and
the next question is given immediately. This switchover of the decision
model can take place either with or without the counter being reset to
zero. If the value v is exceeded again within the time interval the
interval can be extended.
Needless to say, if the program section 56 is not the first program section
in the dialogue, then a block, such as block 37, is switched between
blocks 31 and 32 so that the user can protest if a command n-1 is wrongly
recognized in the preceding program section and the question n-1 can be
asked again.
The branch 52 need not necessarily contain counter block 40. If the branch
52 does not contain block 40 then this branch runs directly via block 53
to block 31--i.e. not via 43--to repeat the nth question.
In FIG. 2A the decision model 33 operates in the same way as in FIG. 2, but
has a slight change in the branch 49. In block 50' in branch 49' the
system asks the user to affirm:
"Did you give Amsterdam as a command?" to which the user can reply "Yes" or
"No" in 51'. If the answer is "Yes" the program then runs to block 36 for
the next question from the system. If the answer is "No" the program then
runs via branch 52, possibly via block 40 (if available) to block 53. A
check such as this is very useful since the recognition of only two
commands, namely "Yes" and "No", is very simple. Preferably, block 40 in
branch 52 should not be omitted here. The answer "No" from the user in
block 51' is now regarded as a protest and counted.
The time interval during which the decision model in FIGS. 2 and 2A
proceeds through branch 49 or 49', respectively, after recognition, can
have a fixed value. This means that after the recognition in 33, 33', 33",
. . . of those commands pertaining to the questions n, n+1, n+2, . . .
which fall within the time interval in question, the system gives the more
explicit indication in accordance with blocks 50 or 50' respectively.
Another possibility is that after the decision model switches over to the
more explicit indication (branch 49 or 49', respectively), a count is made
of how many subsequent commands are then correctly recognized. After, for
example, three times a correct recognition of a command the system then
again switches over to the reaction in accordance with branch 57 (in other
words, no further indication of the command which has been recognized). In
this case the time interval is related to the time during which correct
recognition of three (subsequent) commands has taken place.
Yet another possibility is that the more explicit indication of the command
takes place during the relevant program section 56, and possibly
subsequent program sections--for example, also program section 56' (see
FIG. 2). In that case the time interval is related to the length in time
of the relevant program section or sections.
The value v will be chosen greater than or equal to two. On the other hand,
it should not be too great in order to avoid the need to protest too many
times before the decision model is to change. The point is that too many
protests are annoying to a user. A suitable value for v may lie between
two and, say, five.
FIG. 3 shows another method in accordance with the invention. The method
displays a great deal of similarity with the method in accordance with
FIG. 2. The difference here is in the decision model 56 and in the fact
that block 48 from FIG. 2 is omitted here. The decision model here is
equipped in such a way that, depending on the score of a command which has
been recognized, the system selects from a number of reactions. The first
reaction is that in which the system gives the next question immediately
after a recognition, see branch 57 which leads directly to block 36. The
second reaction--branch 49--is that in which the system gives information
about the command which has been recognized. This is in accordance with
the operation of blocks 50 and 51 as already described on the basis of
FIG. 2. The third reaction--branch 49'--is that in which the system reacts
as described by means of blocks 50' and 51' on the basis of FIG. 2A.
It should be added here that the system need not necessarily be able to
give three reactions after recognition. It is also possible, of course,
for the system to select from two reactions after recognition, for
example, in accordance with branch 57 or in accordance with branch 49.
The operation of the decision model 56 is explained in greater detail on
the basis of FIG. 4. It is known from speech recognition theory that a
score is allocated to a recognition indicating the degree of reliability
of the command which has been recognized. This score may be between 0% and
100%. A score axis is plotted in FIG. 4A. If the score falls in the
interval between 0 and x % then it is decided that the command has not
been recognized. If the scores for all commands fall in this range then
obviously no command whatsoever has been recognized. The program then runs
back via branch 34. We now assume that the command recognition unit has
ultimately come to the conclusion that a particular command has been
recognized. The way in which the command recognition unit has arrived at
that conclusion is not important here. If the score for the command which
has been recognized falls in the interval between x and y %, then the
system gives the reaction in accordance with block 50'--in branch
49'--whereupon the recognition unit is again switched on for the reaction
of an affirmative or a negative answer from the user.
If the score for the command which has been recognized falls in the
interval between y and z %, then the program follows branch 49. If the
score falls in the intervals between z % and 100% then the program follows
branch 57--in other words, it is then impossible for the user to check
immediately on the command which has been recognized.
The score axis is also plotted in FIG. 4b, but now for the situation in
which the decision model 56 is adjusted as a consequence of the number of
protests exceeding the value v. The boundaries y and z from FIG. 4a are
shifted to higher score values.
For the purpose of adjustment, a command which has been recognized with a
score s.sub.1 will lead to a reaction of the system that runs via branch
57 and, after adjustment of the decision model, will lead to a reaction
via branch 49. Similarly, a command which has been recognized with a score
s.sub.2 will lead to a reaction in accordance with branch 49 (before
adjustment) or branch 49' (after adjustment), respectively.
If the user gives a negative answer in block 51' the program goes on via
branch 52'. There are again two possibilities. Either the program runs via
block 53 directly to block 31 in accordance with the branch 58 given in
dashed lines, or the program runs to block 40 and then via block 53 and
block 43 also returns to block 31.
The decision model 56 can also be modified in a different way in yet
another example of an embodiment of the flow chart of FIG. 3. For example,
it is also possible that, after the counter status m has exceeded the
value v, the decision model 56 gives only a reaction in accordance with
circuit 49' during the time interval, after a recognition of a command. At
the end of the time interval the decision model then returns to the
possible reactions in accordance with circuits 57, 49 and 49' depending on
the score, as discussed on the basis of FIG. 4a.
The two decision models 56 as discussed above differ, among other things,
in one important respect from the decision models 33 of FIGS. 2 and 2a.
In the decision models of FIGS. 2 and 2a the program runs during the
interval in question for all the commands recognized via the path (circuit
49 or 49', respectively) of a more explicit indication of the command
recognized. In the decision models 56 as described above that is not the
case. In the decision model 56 as described on the basis of FIGS. 4a and b
there are commands (namely those commands with a scoring percentage lying
between x and y, between y' and z and between z' and 100) for which the
change in the decision model to the situation of FIG. 4b as a consequence
of the fact that m has exceeded the value v, has no effect on the reaction
of the system.
Even if the decision model 56 changes in such a way that after recognition
of a command only a reaction in accordance with circuit 49' takes place,
the system's reaction to commands which are recognized with a scoring
percentage lying between x and y remains the same. This is why it was
stated at an earlier stage that in at least a number of the cases the
system chooses a reaction in step (c) in the form of a more explicit
indication of the command which has been recognized. For one example of an
embodiment of the decision model 56 in FIG. 3 these are therefore the
cases in which the scoring percentage of the command which has been
recognized lies between y and y' and between z and z'. For the other
example of an embodiment in FIG. 3 these are therefore those cases in
which the scoring percentages lie between y and 100.
FIG. 5 shows an example of an embodiment of an interactive system for the
implementation of the method in accordance with the invention. The system
contains a question unit 60 for asking the user of the system a question
in the form of speech. For this purpose, the output 61 of the question
unit 60 is coupled to a loudspeaker 62. In addition, there is a command
recognition unit 63 for the recognition of a command--in command receiver
block 64--and containing a decision unit 65. An activation unit 66 has a
control input 67 coupled via the line 68 to an output 69 of the command
recognition unit 63. An input 70 of the command recognition unit 63 is
coupled to a microphone 71.
The microphone 71 is additionally coupled to a protest recognition unit 72
for the recognition of a protest supplied to the system by the user if a
command is wrongly recognized by the system. This therefore involves an
acoustic protest. At the moment when a protest occurs the protest
recognition unit 72 supplies a counter pulse to its output 73, which
counter pulse is supplied to a count input 74 of a counter 75. The output
76 of this counter 75 is coupled to a first input 77 of a comparison
circuit (comparator) 78. The value v is supplied to the second input 79.
The content m of the counter 75 is supplied to the first input 77. The
output 80 is coupled to a control input 81 of the decision unit 65. The
comparator 78 produces a control signal as its output 80 if the value m
exceeds the value v.
In addition, there is also a central control unit 82 which controls the
various components of the system. For example, there is a control line 83
for driving the question unit 60. Following a control signal over this
line 83 the question unit 60 will present a question to the user in the
form of speech via the loudspeaker 62. There is a line 84 for next
switching on the command recognition unit 63 (and later also switching it
off again). There is also a control line 85 for switching on the protest
recognition unit 72 at the right moment so that a protest can be
recognized. There is a line 86 from the output 80 to the central control
unit 82, so that the central unit knows when m exceeds the value v. A
control line 87 to the counter 75 resets it to zero as soon as a signal
occurs on the line 86. The line 68 is also coupled to the central unit 82.
The operation of the arrangement shown in FIG. 5 for enabling the method of
FIG. 2 to be implemented is now explained as follows. If m.ltoreq.v then
the program shown in FIG. 2 runs from the block 33 either via the branch
| | |
