Cm,p,u.
Biol. Med. Vol. 22, No. 6. pp. 437-441,
0010-4825/92 ss.oo+ .co @ 1992 Pergamon Press Ltd
1992
Printed in Great Britain
A FITNESS ANALYSIS SYSTEM WITH AN INTELLIGENT INTERFACE ALFIO V. PARISI
and GRAHAM D. ALLEN
School of Applied Science, University of Southern Queensland, Toowoomba, Qld 4350, Australia (Received 29 July 1991; in revised form 17 March 1992; received for publication 1 April 1992)
Abstract-This paper describes the development of a system with an intelligent interface for analysis of physiological correlates of athletes’ physical performance capacities. The system improves the interface between the physiologist and the coach and provides scientific information in a systematic and coherent fashion. The recommendations provided are based on the results of a series of physiological tests. The implementation of the system is described with emphasis placed on recognition of the internal structure of the knowledge, independence from a particular shell, design for future expansion and maintenance and the integration with existing information resources. Fitness analysis Physiological tests
Physical performance Prolog
Intelligent interface
INTRODUCTION
The development and success of the Australian Institute of Sport attests to the need for a high level of sophistication in modern coaching and training techniques designed to enhance athletic performance. The effects of particular training schedules on athletes must be monitored closely and frequently. However, one of the problems in physiological testing is lack of feedback of the results of the physiological tests and their implications for performance and training. The purpose of this study was to develop a fitness analysis system with an intelligent interface, EXFIT, to bridge the gap between the scientist and the coach and facilitate the provision of scientific information in a systematic and coherent fashion. EXFIT uses the results of a series of physiological assessments of the athlete to provide generalized recommendations which the coaches may utilize in developing athlete-specific training schedules. The application of such a system in performance analysis has many potential benefits including: 0 Enhancing the interface between the results of scientific assessment and the speed of their application to the field situation. 0 Describing the physiological changes that should be noted in the athlete undertaking the recommended activities. Systems with intelligent interfaces applied to fitness analysis and physical performance appear to be rare. Vickers and Kingston [l] described an expert system for ice hockey consisting of a laser videodisc controlled by an expert shell. Sainsbury et al. [2] outlined an on-line microcomputer system for monitoring of physiological variables during an aerobic power analysis. Owens [3] developed a Commodore 64 microcomputer system for monitoring heart-rate, and for cardiovascular, muscular and body fat analyses. EXFIT has been extended beyond these systems to provide coaches and athletes with systematic and logical recommendations based on physiological assessments. METHODS The development of EXFIT required the knowledge of one or more experts to be captured and stored in such a way that the knowledge may be utilized to make 437
A. V.PARISIandG. D. ALLEN
438
MAINTAIN USER INFORMATION INTERFACE KNOWLEDGE Fig. 1. Structure of EXFIT.
recommendations. Debenham [4] suggests that the common errors in using “standard” methods of knowledge representation [5] include: (a) systems not designed for maintenance; (b) knowledge committed to a particular shell; (c) knowledge collected with little regard for its internal structure; and (d) knowledge based systems designed independently of any existing information. PROCESSING
RESOURCES
To avoid the first two of these errors, EXFIT was structured to contain data, information and knowledge. Debenham [4] states that data are the objects stored, information consists of the relationships and associations between the items of data, and knowledge is the relationships between the items of information and/or the items of data. Rather than being stored as directly executable code, the data and information are stored in a conventional database using the knowledge dictionary concept outlined by Jansen and Compton [6], Jansen [7] and Debenham [8]. As indicated in Fig. 1, the user interface of EXFIT is independent of the data, information and knowledge, and any updates are independent of the user interface. MAINTAIN is a program written to enable the domain expert to modify or add to the knowledge base by taking into account the interrelationships between the entries and reducing the tedium and errors associated with manual updating. To avoid the third of Debenham’s [4] errors, that of knowledge being gathered without regard for its internal structure, the recommendations within EXFIT were developed following interviews with the domain expert. This resulted in the development of an initial prototype which was subject to analysis, revision and refinement by the domain expert and three professional coaches. The structure, depicted in Fig. 2, requires the user to select those particular analyses to be applied on the athlete. For each analysis, EXFIT uses a forward chaining technique [9] to infer a set of tests to be applied, with the user entering the test results which have been collected either manually or automatically. From these, EXFIT concludes a possible diagnosis and generates a set of recommendations with aims, associated activities and follow-up procedures (including the time within which repeat tests should be applied), and appropriate explanations and warnings. The explanations outline to the coach and athlete any physiological changes that may be anticipated should the recommendations be adopted and applied by the coach. Warnings detail any particular caution that must be addressed. An example of each is provided in Fig. 3 which is a sample output of EXFIT following the analysis of the aerobic status of muscle. EXFIT has been designed for application within a human performance laboratory which has specific orientation to the testing of athletes from a variety of sports and with a range of performance capacities. As such it has addressed the fourth of Debenham’s [4] errors, in that the design and development of EXFIT has been dedicated to the enhancement of existing computer software for physiological assessment. Presently the series of analyses available in EXFIT are: peak five second anaerobic power; total anaerobic power; aerobic power; onset of blood lactate accumulation; maximum blood lactate; blood analysis; muscle structure (fibre type); metabolic status of muscle-aerobic and anaerobic capacities.
A fitness analysis system
439
5-
TESTS
DIAGNOSIS
Fig. 2. Internal structure of knowledge.
To further enhance the interface with the field application, EXFIT provides two options based on graphical displays, one which contrasts the athlete’s current test results with those of elite counterparts in the same sport, the other which provides an indication of the changes in the athlete’s physiological capacities over time. Presently these options are restricted to the aerobic and anaerobic power analyses. The complete system runs on an IBM-PC or compatible. An efficient and flexible user interface guides the user by a series of simple screenforms and window/menu interfaces. SUMMARY One of the problems in physiological assessment of the athlete is lack of feedback of the results of the physiological tests and their implications for performance and training. The purpose of this study was to develop a fitness analysis system with an intelligent interface, EXFIT to bridge the gap between the scientist and the coach and facilitate the provision of scientific information in a systematic and coherent fashion. EXFIT uses the results of a series of physiological assessments of the athlete to provide generalized recommendations which coaches may utilize in developing athlete-specific training schedules. The system has been designed and developed for an IBM-PC or compatible, and is intended to enhance the interface between the results of scientific assessment and the development and application of appropriate programs in the field situation. The knowledge in EXFIT has been structured to consist of a series of analyses that may be carried out on athletes. Each particular analysis may require one or more tests to be performed. Using the test results, EXFIT concludes a possible diagnosis for which EXFIT provides one or more aims with associated activities and follow-up procedures. For each activity, EXFIT outlines appropriate explanations and warnings. The knowledge is represented in a form that is not committed to a particular shell, takes into account the internal structure of the knowledge, allows ease of maintenance and future expansion and is integrated with existing information processing resources. The analyses presently available in EXFIT are: anaerobic power (total and peak five second), aerobic power, onset of blood lactate accumulation, maximum blood lactate, blood, muscle structure (fibre type), metabolic status of muscle-aerobic and anaerobic. The test data from these analyses are collected either automatically or manually.
440
A. V. PARISLand G. D. ALLEN
RECOMMENDATIONS TO IMPROVE PERFORMANCE AIM: Increase glycogen level Increase the level of aerobic markers in muscle ACTIVITY: Increase endurance training (intensity and duration) Increase carbohydrate content of diet Include 60 to 240 second surges of exercise in an endurance training programme EXPLANATION: An increase in the stores of muscle glycogen can improve endurance performance. Muscle glycogen is a primary source of energy during prolonged aerobic activity Cardiorespiratory function will improve significantly Skeletal muscle will be affected in many ways by endurance training (1) The ability to oxidise carbohydrates and fats will increase (2) myoglobin concentration in the muscle will increase (3) muscle stores of the fuels glycogen and triglycerides (energy sources) will increase
WARNING: Do not increase your overall caloric load FOLLOW
UP PROCEDURE:
Return in six weeks REASONING
- test
results used to provide recommendations are:
normal moderately trained
Glycogen Level Aer Markers - SDH,CS,CYT-OX
Fig. 3. Sample output of EXFIT for the aerobic status of the muscle analysis.
CONCLUSION A fitness analysis system with an intelligent interface has been described. It improves the feedback from the results of physiological assessments of the athlete to the coach. The applicability of knowledge based systems has been proven in a new field. The structure of the real knowledge has been recognized and the internal knowledge representation uses this structure. The knowledge is represented in a form that is not committed to a particular shell. The knowledge is completely independent of the user interface and is designed to allow ease of maintenance. The system has been integrated with the data acquisition software in a human performance laboratory. In this fashion, the design of the knowledge base provides a thorough, practical testing of the theory of knowledge base design proposed by Debenham [4]. Acknowledgements-The authors would like to thank Ron Matthews in the mechanical workshop of the School of Applied Science, USQ for hisassistance.
REFERENCES 1. J. S. Vickers and G. E. Kingston, Modelling the master coach: Building an expert system for coaching, Proc. ht.
Conf. Comput. Assist. Learn. Post-second. Educat. Calgary, Canada (1987).
A fitness analysis system
441
2. D. A. Sainsbury, C. J. Gore, R. T. Withers and A. H. Ilsley, An online microcomputer program for the monitoring of physiological variables during rest and exercise, Compur. Viol. Med. 18, 17-24 (1988). 3. R. Owens, FITSCAN, A computer monitoring and analysis of cardiovascular performance, ACM Thirteenth unn. Comput. Sci. Conf. New York (1985). 4. J. K. Debenham, Managing expert systems: four key issues, Proc. Fifth Australian Conf. Appl. Expert Syst. Sydney, 163-188 (1989). 5. B. Jansen and P. Compton, The knowiedge dictionary: storing different knowledge representations, Proc. Fifth Australian Conf. Appl. Expert Syst. Sydney, 143-162 (1989). 6. B. Jansen and P. Compton, The knowledge dictionary: A relational tool for the maintenance of expert systems, CSIRO Division of Information Technology Technical Report TR-FC-88-08 (1988). 7. B Jansen, The knowledge dictionary: integrating a knowledge base with a data store, CSZRO Division of Information Technology Technical Report TR-FC-88-02 (1988). 8. J. K. Debenham, Knowledge base design, Austral. Comput. J. 17,42-48 (1985). 9. C. Townsend, Mastering Expert Systems with Turbo Prolog. pp. 67 and 139-141. Howard W. Sams & Co., Indianapolis (1986). About the Anther-ALFIO V. PARIS] received a B.Sc.(Hons) degree in Physics from the James Cook University in 1981. From 1981 to 1987, Mr Parisi worked as a geophysicist with the Bureau of Mineral Resources which included a large proportion of time on marine geophysical surveys. Currently he is a Physics lecturer in the School of Applied Science at the University of Southern Queensland and has recently submitted a thesis for a M.App.Sc.
About the Author-GRAHAM DONALD ALLEN attended Sydney Teacher’s College, Sydney University, the University of British Columbia, and received his Ph.D. from Washington State University in 1975. He joined the faculty of the Darling Downs Institute of Advanced Education in 1975, and is currently Associate Professor in the Schools of Applied Science and Education at the University of Southern Queensland. His current research interests include muscle biochemistry, muscle damage with exercise, recovery processes following exercise and sport science. He is Head of the Centre for the Assessment of Human Performance, and is a Fellow of the Australian Sports Medicine Federation.
Cen 22-6-E
Biol. Med. Vol. 22, No. 6. pp. 437-441,
0010-4825/92 ss.oo+ .co @ 1992 Pergamon Press Ltd
1992
Printed in Great Britain
A FITNESS ANALYSIS SYSTEM WITH AN INTELLIGENT INTERFACE ALFIO V. PARISI
and GRAHAM D. ALLEN
School of Applied Science, University of Southern Queensland, Toowoomba, Qld 4350, Australia (Received 29 July 1991; in revised form 17 March 1992; received for publication 1 April 1992)
Abstract-This paper describes the development of a system with an intelligent interface for analysis of physiological correlates of athletes’ physical performance capacities. The system improves the interface between the physiologist and the coach and provides scientific information in a systematic and coherent fashion. The recommendations provided are based on the results of a series of physiological tests. The implementation of the system is described with emphasis placed on recognition of the internal structure of the knowledge, independence from a particular shell, design for future expansion and maintenance and the integration with existing information resources. Fitness analysis Physiological tests
Physical performance Prolog
Intelligent interface
INTRODUCTION
The development and success of the Australian Institute of Sport attests to the need for a high level of sophistication in modern coaching and training techniques designed to enhance athletic performance. The effects of particular training schedules on athletes must be monitored closely and frequently. However, one of the problems in physiological testing is lack of feedback of the results of the physiological tests and their implications for performance and training. The purpose of this study was to develop a fitness analysis system with an intelligent interface, EXFIT, to bridge the gap between the scientist and the coach and facilitate the provision of scientific information in a systematic and coherent fashion. EXFIT uses the results of a series of physiological assessments of the athlete to provide generalized recommendations which the coaches may utilize in developing athlete-specific training schedules. The application of such a system in performance analysis has many potential benefits including: 0 Enhancing the interface between the results of scientific assessment and the speed of their application to the field situation. 0 Describing the physiological changes that should be noted in the athlete undertaking the recommended activities. Systems with intelligent interfaces applied to fitness analysis and physical performance appear to be rare. Vickers and Kingston [l] described an expert system for ice hockey consisting of a laser videodisc controlled by an expert shell. Sainsbury et al. [2] outlined an on-line microcomputer system for monitoring of physiological variables during an aerobic power analysis. Owens [3] developed a Commodore 64 microcomputer system for monitoring heart-rate, and for cardiovascular, muscular and body fat analyses. EXFIT has been extended beyond these systems to provide coaches and athletes with systematic and logical recommendations based on physiological assessments. METHODS The development of EXFIT required the knowledge of one or more experts to be captured and stored in such a way that the knowledge may be utilized to make 437
A. V.PARISIandG. D. ALLEN
438
MAINTAIN USER INFORMATION INTERFACE KNOWLEDGE Fig. 1. Structure of EXFIT.
recommendations. Debenham [4] suggests that the common errors in using “standard” methods of knowledge representation [5] include: (a) systems not designed for maintenance; (b) knowledge committed to a particular shell; (c) knowledge collected with little regard for its internal structure; and (d) knowledge based systems designed independently of any existing information. PROCESSING
RESOURCES
To avoid the first two of these errors, EXFIT was structured to contain data, information and knowledge. Debenham [4] states that data are the objects stored, information consists of the relationships and associations between the items of data, and knowledge is the relationships between the items of information and/or the items of data. Rather than being stored as directly executable code, the data and information are stored in a conventional database using the knowledge dictionary concept outlined by Jansen and Compton [6], Jansen [7] and Debenham [8]. As indicated in Fig. 1, the user interface of EXFIT is independent of the data, information and knowledge, and any updates are independent of the user interface. MAINTAIN is a program written to enable the domain expert to modify or add to the knowledge base by taking into account the interrelationships between the entries and reducing the tedium and errors associated with manual updating. To avoid the third of Debenham’s [4] errors, that of knowledge being gathered without regard for its internal structure, the recommendations within EXFIT were developed following interviews with the domain expert. This resulted in the development of an initial prototype which was subject to analysis, revision and refinement by the domain expert and three professional coaches. The structure, depicted in Fig. 2, requires the user to select those particular analyses to be applied on the athlete. For each analysis, EXFIT uses a forward chaining technique [9] to infer a set of tests to be applied, with the user entering the test results which have been collected either manually or automatically. From these, EXFIT concludes a possible diagnosis and generates a set of recommendations with aims, associated activities and follow-up procedures (including the time within which repeat tests should be applied), and appropriate explanations and warnings. The explanations outline to the coach and athlete any physiological changes that may be anticipated should the recommendations be adopted and applied by the coach. Warnings detail any particular caution that must be addressed. An example of each is provided in Fig. 3 which is a sample output of EXFIT following the analysis of the aerobic status of muscle. EXFIT has been designed for application within a human performance laboratory which has specific orientation to the testing of athletes from a variety of sports and with a range of performance capacities. As such it has addressed the fourth of Debenham’s [4] errors, in that the design and development of EXFIT has been dedicated to the enhancement of existing computer software for physiological assessment. Presently the series of analyses available in EXFIT are: peak five second anaerobic power; total anaerobic power; aerobic power; onset of blood lactate accumulation; maximum blood lactate; blood analysis; muscle structure (fibre type); metabolic status of muscle-aerobic and anaerobic capacities.
A fitness analysis system
439
5-
TESTS
DIAGNOSIS
Fig. 2. Internal structure of knowledge.
To further enhance the interface with the field application, EXFIT provides two options based on graphical displays, one which contrasts the athlete’s current test results with those of elite counterparts in the same sport, the other which provides an indication of the changes in the athlete’s physiological capacities over time. Presently these options are restricted to the aerobic and anaerobic power analyses. The complete system runs on an IBM-PC or compatible. An efficient and flexible user interface guides the user by a series of simple screenforms and window/menu interfaces. SUMMARY One of the problems in physiological assessment of the athlete is lack of feedback of the results of the physiological tests and their implications for performance and training. The purpose of this study was to develop a fitness analysis system with an intelligent interface, EXFIT to bridge the gap between the scientist and the coach and facilitate the provision of scientific information in a systematic and coherent fashion. EXFIT uses the results of a series of physiological assessments of the athlete to provide generalized recommendations which coaches may utilize in developing athlete-specific training schedules. The system has been designed and developed for an IBM-PC or compatible, and is intended to enhance the interface between the results of scientific assessment and the development and application of appropriate programs in the field situation. The knowledge in EXFIT has been structured to consist of a series of analyses that may be carried out on athletes. Each particular analysis may require one or more tests to be performed. Using the test results, EXFIT concludes a possible diagnosis for which EXFIT provides one or more aims with associated activities and follow-up procedures. For each activity, EXFIT outlines appropriate explanations and warnings. The knowledge is represented in a form that is not committed to a particular shell, takes into account the internal structure of the knowledge, allows ease of maintenance and future expansion and is integrated with existing information processing resources. The analyses presently available in EXFIT are: anaerobic power (total and peak five second), aerobic power, onset of blood lactate accumulation, maximum blood lactate, blood, muscle structure (fibre type), metabolic status of muscle-aerobic and anaerobic. The test data from these analyses are collected either automatically or manually.
440
A. V. PARISLand G. D. ALLEN
RECOMMENDATIONS TO IMPROVE PERFORMANCE AIM: Increase glycogen level Increase the level of aerobic markers in muscle ACTIVITY: Increase endurance training (intensity and duration) Increase carbohydrate content of diet Include 60 to 240 second surges of exercise in an endurance training programme EXPLANATION: An increase in the stores of muscle glycogen can improve endurance performance. Muscle glycogen is a primary source of energy during prolonged aerobic activity Cardiorespiratory function will improve significantly Skeletal muscle will be affected in many ways by endurance training (1) The ability to oxidise carbohydrates and fats will increase (2) myoglobin concentration in the muscle will increase (3) muscle stores of the fuels glycogen and triglycerides (energy sources) will increase
WARNING: Do not increase your overall caloric load FOLLOW
UP PROCEDURE:
Return in six weeks REASONING
- test
results used to provide recommendations are:
normal moderately trained
Glycogen Level Aer Markers - SDH,CS,CYT-OX
Fig. 3. Sample output of EXFIT for the aerobic status of the muscle analysis.
CONCLUSION A fitness analysis system with an intelligent interface has been described. It improves the feedback from the results of physiological assessments of the athlete to the coach. The applicability of knowledge based systems has been proven in a new field. The structure of the real knowledge has been recognized and the internal knowledge representation uses this structure. The knowledge is represented in a form that is not committed to a particular shell. The knowledge is completely independent of the user interface and is designed to allow ease of maintenance. The system has been integrated with the data acquisition software in a human performance laboratory. In this fashion, the design of the knowledge base provides a thorough, practical testing of the theory of knowledge base design proposed by Debenham [4]. Acknowledgements-The authors would like to thank Ron Matthews in the mechanical workshop of the School of Applied Science, USQ for hisassistance.
REFERENCES 1. J. S. Vickers and G. E. Kingston, Modelling the master coach: Building an expert system for coaching, Proc. ht.
Conf. Comput. Assist. Learn. Post-second. Educat. Calgary, Canada (1987).
A fitness analysis system
441
2. D. A. Sainsbury, C. J. Gore, R. T. Withers and A. H. Ilsley, An online microcomputer program for the monitoring of physiological variables during rest and exercise, Compur. Viol. Med. 18, 17-24 (1988). 3. R. Owens, FITSCAN, A computer monitoring and analysis of cardiovascular performance, ACM Thirteenth unn. Comput. Sci. Conf. New York (1985). 4. J. K. Debenham, Managing expert systems: four key issues, Proc. Fifth Australian Conf. Appl. Expert Syst. Sydney, 163-188 (1989). 5. B. Jansen and P. Compton, The knowiedge dictionary: storing different knowledge representations, Proc. Fifth Australian Conf. Appl. Expert Syst. Sydney, 143-162 (1989). 6. B. Jansen and P. Compton, The knowledge dictionary: A relational tool for the maintenance of expert systems, CSIRO Division of Information Technology Technical Report TR-FC-88-08 (1988). 7. B Jansen, The knowledge dictionary: integrating a knowledge base with a data store, CSZRO Division of Information Technology Technical Report TR-FC-88-02 (1988). 8. J. K. Debenham, Knowledge base design, Austral. Comput. J. 17,42-48 (1985). 9. C. Townsend, Mastering Expert Systems with Turbo Prolog. pp. 67 and 139-141. Howard W. Sams & Co., Indianapolis (1986). About the Anther-ALFIO V. PARIS] received a B.Sc.(Hons) degree in Physics from the James Cook University in 1981. From 1981 to 1987, Mr Parisi worked as a geophysicist with the Bureau of Mineral Resources which included a large proportion of time on marine geophysical surveys. Currently he is a Physics lecturer in the School of Applied Science at the University of Southern Queensland and has recently submitted a thesis for a M.App.Sc.
About the Author-GRAHAM DONALD ALLEN attended Sydney Teacher’s College, Sydney University, the University of British Columbia, and received his Ph.D. from Washington State University in 1975. He joined the faculty of the Darling Downs Institute of Advanced Education in 1975, and is currently Associate Professor in the Schools of Applied Science and Education at the University of Southern Queensland. His current research interests include muscle biochemistry, muscle damage with exercise, recovery processes following exercise and sport science. He is Head of the Centre for the Assessment of Human Performance, and is a Fellow of the Australian Sports Medicine Federation.
Cen 22-6-E
