Choosing signal processing algorithms for AQSs should be based on a pre
vious analysis of situations in which the given AQS will be used. It is not sufficient to answer the questions: what sort of signals should be received and what is the suggested sort of substitutive sensor that will be used instead of the injured one. The next question is: in what conditions the signals will
evident that AQSs, in general, cann't substitute natural sensors but in sin
gular and rather specific situations. The next problen arising is thus selec
tion of signal processing tasks that are preferred as being aided by the AQS.
Any task of this sort should be, in addition, characterized by the desired ac
curacy of solution as well as by its admissible response-tine /so as being sol
ved by the AQS in real-tine/. For example, for blind people the following tasks can be specified:
Task characteristic Accuracy Response-time
Detection of large moving objects middle
<<
0,3 sEvaluation of distance to large
ata-tionary objects low < 2,0 s
Detection of objects of given shape
and size middle < 1,0 B
Recognition of indications on displays high < 1,5 8 leading printed texts:
- for a fixed alphabet high
** 20
char/s- for several collections of char
acters middle 15 char/s
Here "low" accuracy means more than 30% of erroneous answers, "middle" ac
curacy - between 10 and 30% and "high" accuracy - less than 10%, however, the numbers has been taken in an arbitrary way*
It is also evident that the last two tasks can be recommended for statio
nary visual AQSs rather while the former ones can be realised by handy, porta
ble AQSs.
Host of the above-mentioned tasks belong to the class of pattern recogni
tion problems and can be solved using the well-known methods based on
determi-* I
niatlc, statistical or structural approach / , L33 /. The real-time signal processing condition leads to some additional technological problems, so as the received optical signals should be stored in an image buff er «memory before processing. The specific technological problems will be not considered here.
In similar way, for deaf and/or deaf-mute people the following signal-pro
Task characteristic Accuracy Response-time Detection of any loud burst of noise
with indication of direction middle < 0,2 s Recognition of typical acoustic sig
nals /doorbell, huAao voice, dog's
bark etc./ middle < 0,5 s
Recognition of selected single words
spoken middle < 0,5 s
Loud reading of short texts introdu
ced through a keyboard high < ■ Displaying in typed form spoken texts middle < 1,0 8
Effective solution of some of the above-mentioned tasks needs using ad
vanced information processing methods. For example, displaying continuous spo
ken texts is a problem that concerns speech analysis on morphological, syntacti
cal and sometimes on semantical levels. The problem needs thus high data proces
sing. rate as well as high RAM's capacity.
All the above-mentioned signal-processing tasks were based on an assump
tion of steady working conditions in which using fixed algorithms and programs is possible. However, the AQSs based on such principle would be rather unflex
ible and in fact uneffective. The working conditions may change due to the chan
ges in the environment the invalid exists and acts as well as due to the chan
ges of his proper needs. In particular, it is necessary to take into account the fact that the abilities of invalids to solve their vital problems develop according to the experience stored, to the health state etc. This means that the AQSs should be flexibly programmed and able to be adapted to changing cir
cumstances*
So as the AQSs perform their signal-processing operations in real time, they are programmed so as to repeat their programs in loops with repetition
time not overpassing the admissible response-time of the AQS. However, in or
der to make the programs adjustable to the current user's demands it is defined up to a certain set of values of parameters controlled by the user. The
parame-i. calculation constants, like: sigsal threshold levels, time-inter
val a duration, frequencies, signal-shape characteristics etc.}
ii. program constants, liket dimensions of arrays, nuaber of repeti- tioas in program loops, subroutine labels etc.;
iii. structural algorithm constants, like labels assigned to optional connections within a general logical schene of a program etc.
The set of admissible values of adjusted parameters determines a sort of a space of control signals for the u s e r of
Mjß.
The parameter values can be set by keys and stored in registers for be i n g used during the program perfomance.From the user's point of view the keyboard states thus form a sort of expres
sions of a manipulation language that can be used for AQS hand-control* How
ever, it is assumed that the user is not able to set his own programs of sig
nal processing, the last being designed by the producer of AQS* It is also clear that the degree of freedom in program adjusting offered to the users is larger in stationary AQSs than in the portable ones*
Future progress in AQSs construction will concern their ergonomic paramo*
ters, reliability, versatility, sensibility and accuracy* The manipulation lan
guage will be "naturalised" so as to m a k e AQS control easier and more flexible*
This will be reached due to wide application in AQSs construction all achieve-e
ments in microelectronic schemes technology, in pattern recognition and artifi
cial intelligence and in construction of si&ial receptors.
Refer «ices
1* R*F* Schmidt, G* Thews. Human Physiology. Springer Verlag, Berlin, 1983*
2. J. Kulikowski. Cybernetycxne uklady roxpoznajqce. PWK, Warszawa, 1972.
3* K.S* Fu* Syntactic methods in pattern recognition. Acad* Press, Hew York, 1974
ON THE PADRE'S IMPLEMENTATION BY PREPROCESSOR TECHNIQUE
Different steps and advantages of preprocessor tech
nique are showed in this paper by way of a concrete imple
mentation of Padre, a programming language especially designed for describing, transforming and interpreting vari
ous classes of Petri nets. We also show that the technique is perfectly applicable on microcomputers.
1. I N I R O D U e i l Q N
It has been exposed in C13 that Padre is an experimen
tal programming language which can describe, transform and interpret different classes of Petri nets - e.g ordinary nets, capacity nets (C23), coloured nets (C33), predicate/transition nets (CAT), ... In this paper, a par
ticular Padre's implementation technique is exhibited.
In this implementation we seek two objectives: develop
ment speed and portability. The first one can be obtained by preprocessor technique which translates a high-level source language into another high-level target language. The seman
tics of the source and target languages are often well- understood, so the translation implies no major difficulties because the existence of their direct semantic equivalence.
For portability objective, we have to choose a language which is universally implemented on almost all computer fam
ilies. And that is the case of the language Pascal.
With these objectives in mind, our preprocessor tech
nique consists in four steps:(i) internal representation generating of different types with pertinent informations for future uses, (ii) translating all Padre program into a semantically equivalent Pascal program, (iii) generating useful routines in prevision of user's needs, and (iv) sub
mitting the outputs of preceding steps to a Pascal compiler which completes the translation. The preprocessor design includes of course all major and well-known algorithms of compiling processus (C53) - i.e lexical and syntactical analysis, semantic analysis, error recovery, ... - except
perhaps that the generated object is coded in high-level language.
The paper is organized in three parts: in the first one, we recall some basic constructs of the language Padre, the second part describes our implementation technique, and in the third part we show that this technique is perfectly languages in the description, transformations and interpre
tations of various Petri net classes, we have designed Padre
element_decl = "element" elem_item -C";" elem_item>
elem_item = elem_id -C" , " elem_id> ":" elem_struct elem_struct = "transition" I "simple place" I
"colored place of" int_subrange_id connection_clause = "connection"
trans_connect CinterfaceU {"I"trans_connect Einterface]>
The trans_connect has the following simple syntax trans_connect = trans_id "E"
E i n _ c 1a u s e ";"3 C g u a r d _ c l a u s e " ; " 3 C o u t _ c 1ause]
"3"
in which and for each transition identifier trans_id:
- in_clause specifies all its in_places and its associated