WO2002095714A2

WO2002095714A2 - Method and device for controlling the posture or movement of a person

Info

Publication number: WO2002095714A2
Application number: PCT/EP2002/005525
Authority: WO
Inventors: Jürgen Löschinger
Original assignee: Loeschinger Juergen
Priority date: 2001-05-18
Filing date: 2002-05-18
Publication date: 2002-11-28
Also published as: AU2002338767A1; EP1395968A2; DE10124242A1; WO2002095714A3

Abstract

The invention relates to a method for controlling the posture and/or movement of at least one person with the aid of a training program. The inventive method comprises the following steps: a) data relevant to posture is detected with the aid of sensor devices; b) the data detected by the sensor devices is synchronized and assembled; c) the synchronized and assembled data is filtered and normalized; d) the filtered and normalized data is compared with reference data in order to determine correlation factors and/or examine the occurrence of characteristic features such as extreme values or exceeding threshold values or falling short thereof; e) objects are defined with the aid of the correlation factors and/or occurring characteristic features, said objects comprising a set of conditions and a probability value derived therefrom; f) the objects are processed in an expert databank and messages are outputted depending on the fulfillment of one or several conditions and the actual state of the training program and the subsequent progression of the training program is controlled.

Description

Method and device for checking the posture or movement of a person

description

The invention relates to a method for controlling posture or movement of a person with the aid of a training program. The invention also relates to an apparatus for performing the method.

Methods and devices of the type mentioned are already known. They are used, for example, to learn complex movements, such as dancing, yoga, aerobics, gymnastics, golf, tennis, karate, playing musical instruments and the like. Such complex movement sequences are learned in a known manner with the help of a course program mediation, for example by a personal teacher, or in the form of instructions in books, video programs or the like. Sports, such as golf or tennis, include a combination of many different movements, which can mostly be practiced on their own, but the complex process is another challenge. This complex interaction is typically not learned at once, but in different steps that build on one another. Which steps are involved depends on the one hand on the talent and ability of the practitioner, and on the other hand on the style of the trainer. It is also possible to carry out complex movements without a learn personal trainers, for example, from books or video programs. In the case of books and video programs in particular, errors or inaccuracies frequently occur in the movement execution by the person, so that the learning success can be called into question, and possibly health damage to the person can also be caused due to incorrect loads. A teacher is particularly limited in the possibility of a correct and complete movement analysis in the case of complex and fast movement sequences. Wrong loads on the moving person often only show up afterwards, i.e. after completing the exercise, for example by means of a video recording or using treadmill analyzes by experts. A personal trainer is often relatively expensive and is usually not always available. Even during everyday movements, such as walking or running, there can often be poor physical learning or acquired through other circumstances, which can lead to long-term health problems. Furthermore, such harmful effects due to static postural damage (motionless or poor body positions) of a person can be caused.

It is an object of the invention to provide a method and a corresponding device of the type mentioned at the outset which allow a person to be reliably and quickly checked for posture and movement. To achieve the object, a method for checking the posture of a person is proposed, which has the features of claim 1.

The method for checking the posture and / or movement of a person with the aid of a training program has the following procedural steps:

a) with the help of several sensor devices, posture-relevant data (posture data) are recorded; The acquisition of the data is preferably timed;

b) the data recorded by the sensor devices are put together and in particular synchronized; the data is synchronized, for example, by a control device which outputs a synchronization signal per time unit (for example every 10 Hz); the data received by the various sensor devices are thus coordinated with one another in time; when assembling, the synchronized data are related to each other in time;

c) the composite and in particular synchronized data are filtered and in particular standardized and exempted; When filtering, spatial proximity relationships between the sensors and temporal references are taken into account; filtering serves to smooth the data structure and, for example, to eliminate certain frequencies or around certain ones Reinforce features such as edges;

d) the filtered and standardized data are compared with reference data in order to determine correlation factors and / or are examined for the occurrence of characteristic features such as extreme values or the exceeding or falling below threshold values; depending on the structure and complexity of the data relevant to posture, a comparison with existing reference data or an analysis of the data is preferable, usually a combination of both methods;

e) objects are defined with the aid of the correlation factors and / or the characteristic features found during the examination, each comprising a set of conditions and a probability value derived therefrom; the size of the probability values depends, for example, on the probability of the associated characteristic feature occurring;

f) the objects are processed in an expert database and, depending on the fulfillment of one or more conditions and the current status of the training program, messages are output and the further course of the training program is controlled.

Compared to the known prior art, the method has the advantage that, depending on the person's reaction is promptly given information in the form of messages to improve posture and / or movement. A visual comparison between the current posture and / or movement and an ideal posture and / or movement by the person is not necessary. With the help of the method, the person is guided through the training program without the person having to resort to a human trainer or expert or an appropriate manual. Previous systems were only geared towards isolated movements. The method according to the invention includes a coherent training program. Depending on the reaction of the person practicing, the next step in the process is automatically selected. The process reflects the knowledge and style of a human trainer. Default values are stored in the expert database in the form of reference data, which are used to compare the parameters that are determined from the person's posture. In addition, the expert database contains the training program, which automatically adjusts the training process based on the attitude of the person and their reaction to the information provided.

The invention can also be understood as two hierarchically nested control loops containing one person, the inner control loop (bio-feedback) of the person receiving feedback about the deviation of the posture or movement from an ideal posture or movement (or about their achievement) and the external control loop (MetaFeedback) the independent continuation of the trai- program based on the person's reaction to the feedback from the internal control loop. The training program can thus teach the person movement sequences in individual learning steps, which are adapted to the user, and thus effectively enable the attainment of the superordinate learning goal, namely the learning of complex sequences of movement sequences.

It is advantageously possible to design the expert database in such a way that it contains an internal representation of the user either implicitly or explicitly, for example his training level, his training progress, his daily state, tiredness and / or his state of mind. In this case, reaching a higher training level not only has the effect of having to correspond more exactly to the ideal of a movement sequence, that is to say, receive a corresponding correction feedback even with a small deviation, but also essentially forwarding to areas of the training program, the new, more difficult one Can train and monitor movements. Conversely, the training program can give instructions on less difficult exercises in the event of poor daily health or looming tiredness. This can reduce the risk of injury to the practitioner.

The training program or learning program consists of individual training steps, which can be implemented in the form of a sequence program with instructions and jump commands in a device. For storing or representing the current As of now, a device is required that can be changed by the training program itself, hereinafter referred to as the training program counter. The training program can be defined in its own high-level language based on a programming language and can be executed by an internal interpreter using a training program counter. The course of the training program is guaranteed by the training program counter, which designates the point at which the program is currently located. The training program counter can cause a conditional or unconditional jump in the program flow.

Complex patterns in the input data stream are recognized in the expert database and trigger a corresponding reaction. Depending on the current status of the system, a decision is made as to how to proceed. It is therefore taken into account how the system reacts to the circumstances. In addition to identifying complex patterns, changing the internal state is an important part of the present expert system. This corresponds to a changed position in a state diagram or, for example, a changed position in a table of exercise instructions. An essential feature of the expert system is that the response to the same input data can differ depending on the state of the system. The expert system is used to specifically guide a user through a training program. What the reaction to a user's posture or movement looks like and how to proceed is at the discretion of the human expert who developed the training program. All of the expert knowledge can be implemented as software in the form of an expert database. This makes it possible to run different courses without changing the hardware. The expert system can be used, for example, to recognize movement sequences on the basis of complex patterns of print data, in particular of both feet. The expert system can also be used to run a health and / or fitness check program or, for example, a program to determine the fit of shoes.

The system described is capable of learning and thus adapts to the user. On the one hand, this ability to learn is implicitly built in through the use of the expert system (training program). Depending on the user's reaction to the instructions, the system continues at various points in the learning or training program. At which point exactly depends on the creator (experts) of the actual training program and the user himself. The current status of the training program is saved permanently.

Another type of learning ability is to keep statistics on the user's reaction to the instructions. In an extension of this concept, a special training program can be started, which specifies certain exercises or behaviors in order to get to know the user. NEN. Special instructions can thus be used to evaluate the training status of the user. The reaction to this is evaluated and, together with the statistical data, influences the further course of the training. Another type of learning ability is the consideration of different training levels. Depending on the user and training success, the current training level changes, which automatically leads to a change in the instructions generated.

The training status of the user is therefore characterized and taken into account. Another type of adaptation or learning ability of the system is that the instructions generated to the user can be subjected to a plausibility check and success check before the actual output. This not only prevents contradicting instructions, but also makes it possible, for example, to respond to the user in the language selection. For example, if the user does not respond to an instruction, it can be assumed that the user does not want any instructions at the moment and, for example, wants to rest. In addition, statistical data on the success or failure of certain instructions are determined. This data is in turn used to adapt the output to the user.

The method according to the invention can be used in the field of occupational medicine / safety. To do this, it records the movement sequence, recognizes harmful conditions in good time and gives provide timely instructions for balancing exercises or breaks. It also monitors the effectiveness of passive protective measures such as mats, shoes, insoles etc.

The method can be used to determine a factor associated with the frequency of the movement, with which, for example, the frequency of movement of children can be monitored. The procedure can also be used to check the fit of shoes over a period of hours or days.

A preferred exemplary embodiment of the method is characterized in that the messages are output to the person immediately after the associated conditions have been met, in particular in real time. The person and the sensor devices are, so to speak, parts of a closed control loop which enables a direct influence on the posture or movement of the person.

Another preferred exemplary embodiment of the method is characterized in that the training program takes into account the person's reactions to the messages that are output. This feedback function frees the user from thinking, so to speak.

Another preferred exemplary embodiment of the method is characterized in that the expert database comprises a learning program which has a large number of lines and whose execution is ensured by a counter which stores the Line that the tutorial is on.

Another preferred exemplary embodiment of the method is characterized in that each line of the learning program (training program) comprises an instruction block, a control block and a branching block. The instruction block defines the type of instruction that is issued, for example, via a language unit. The identifiable possible reactions of the practitioner to the instruction are defined in the reference block. The branch block contains a conditional or unconditional jump address for each entry in the control block. This makes it possible to react specifically to the reaction of the practitioner taking into account the current position in the program and to continue at a suitable point. The introduction of conditional jumps makes it possible to influence the program flow by means of additional internal parameters (level of ability, timing, physical condition, etc.). The process is designed in such a way that the competence of the human expert can be recognized in its entirety.

The posture data are preferably determined by means of an optical and / or acoustic recording unit and / or by means of sensors. This enables a quick and reliable posture control of a person without necessarily having to resort to a human expert. According to a preferred embodiment variant, the posture data determined are analyzed in the data processing unit for the subsequent derivation of parameters which are compared with data in the expert database in order to create at least one posture-relevant information for the person. The method can advantageously be used in a versatile manner for controlling a person's posture by means of a suitable choice of an expert database.

A posture reaction to information is preferably used as a criterion for person assessment and for the creation of an adapted information creation. Thus, based on how quickly or how well a corrective postural response has been carried out by a person, knowledge about the respective mental and physical state of the same person can be gained. For example, knowledge can be gained in this way about the ability to coordinate or about the training status of the corresponding person. As a result, the learning program can automatically adjust to the individual circumstances of a person and thus avoid undesirably overwhelming the same person, for example by automatically initiating an interruption in the exercise.

To achieve the object, a device with the features of claim 9 is also proposed, which is characterized in that an expert database which is used for maintaining data evaluation-serving electronic data processing unit is provided, which is operatively connected to the recording unit and to an information output unit serving to transmit at least one posture-relevant information to the person. Here, posture is understood to mean a particular body position of the person in a static position (still) or also during a dynamic movement thereof. Static body positions and body movements of a person can thus be controlled reliably and quickly, the person himself being part of a closed control loop. Depending on the expert database, different posture controls can be carried out, for example for exercises in different sports. It is now advantageously possible to ensure an automated and individualized posture control of a person by means of a device. The information to be transmitted to the person can be, for example, an attitude-related assessment or a movement instruction. Furthermore, it is conceivable that the person is not subjected to a posture control directly, but, for example, an exercise device held by the person (tennis racket, golf club or the like) is checked with regard to its positioning and / or movement for indirect posture control of the person using the exercise device.

In a preferred embodiment, an acoustic recording unit is additionally provided, which is operatively connected to the data processing unit for sound data evaluation. The acute The static recording unit can be a microphone, for example, by means of which sound data can be recorded. This enables, for example, sound data and posture-controlled learning of a musical instrument using a device. The sound data evaluation can also be used indirectly for posture control.

The recording unit advantageously contains at least one sensor for recording posture-relevant data. Sensors are particularly suitable for the reliable and quick acquisition of posture-relevant data. In addition, they can be operatively connected to the electronic data processing unit in a relatively simple manner.

The sensor can be designed as a pressure sensor or as an acceleration sensor. It is thus possible to provide a plurality of different types of sensors for recording posture-relevant data. Pressure sensors can, for example, be arranged in a person's shoe sole to record data relating to the body stress that is occurring in each case. Suitable are, for example, FSR (ForceSensingResitor) sensors, which consist of three thin polymer films and change their electrical resistance depending on the force applied to their surface. They are particularly suitable due to their relatively small installation volume, their long service life and their low purchase costs. The sensors can also be integrated or glued into a plastic measuring sole. The sensors can be glued to the measuring base and preferably be perforated to ensure ventilation of the measuring sole and the removal of moisture. With the same film, a tactile feedback to the person can be created in the form of vibration. It is therefore possible to use the same device, such as a piezo film, to record posture-relevant data and output feedback.

For example, a plurality of pressure sensors can be arranged as a matrix and can be operatively connected to a microcontroller by means of an A / D converter. The individual sensors of the sensor matrix are preferably controlled in a time-coded manner. In a typical application, for example, 9 to 16 pressure sensors are used. With a temporally targeted resolution of 10 ms, there is approximately 0.1 to 1 ms time to record the posture data of the individual sensors and convert them to analog / digital. The conversion time is sufficient to use common, inexpensive microcontroller systems. In this way it is possible to obtain pressure distribution data by means of a plurality of pressure sensors which, based on knowledge of the body weight and the shoe size of the person, allow conclusions to be drawn about the so-called "body weight pressure".

A piezo film can also be used to measure the pressure distribution.

When placing an object equipped with a sensor on another (for example when placing a foot on the ground) or when two objects rub against each other (for example, a ski over snow), structure-borne noise is generated that can be detected by the sensors. Structure-borne noise is used on the one hand to identify the surface, for example the type of ground when running, the snow conditions when skiing, or to provide information on the speed of an object (for example a ski). On the other hand, structure-borne noise can be used to obtain information about the height and the course of the forces acting when two objects collide (for example, a foot on the floor). The structure-borne noise can therefore be used to analyze the movement of the person.

To measure environmental influences, which can influence the reaction of the expert system, sensors for measuring the temperature, the air pressure, the humidity etc. are advantageously used. In addition, the user's personal data, such as pulse, blood pressure, oxygen content in the blood, skin resistance, etc., can be recorded and taken into account in the expert system. For this purpose, sensors are preferably used to record pulse, blood pressure, oxygen in the blood, skin resistance, etc.

Inclination sensors are also preferably used to determine, for example, the inclination of the person's pelvic and shoulder girdles. In addition, acceleration sensors can be used to completely record the movement of the person. According to a possible embodiment, the recording unit has at least one video camera. It is thus possible to carry out an optical posture control of a person using an alternative or additional video camera with respect to sensors. If necessary, two or more video cameras, which are arranged at a fixed distance from one another, can also be provided for the stereo-optical recording of posture data or movement data of a person. Traditional video cameras that generate analog video data or inexpensive web cameras can advantageously be used.

According to a preferred embodiment, at least one optical marking element is provided for attachment to the person. Such optical marking elements are attached to suitable (movement-relevant) points on the person and serve to facilitate data evaluation and thus also to reduce the required computing power. The markings are preferably designed in such a way that they can be clearly identified even if only parts of them are visible. For example, markings of different diameters, different distances from the video camera, different colors and / or different flashing frequencies of LEDs are used.

The information output unit is advantageously designed as a transmitter for optical and / or acoustic and / or tactile signals. Optical signals can be emitted, for example, by means of LEDs, while acoustic signals can be generated by means of a buzzer, for example, or are preferably natural speech signals. Tactile signals can be emitted in the form of a vibration, for example. Natural speech can be delivered using one or more traditional loudspeakers. A combination of different signal forms is also possible in order to be able to transmit fast and reliable posture-relevant information to the person.

The data processing unit advantageously contains a data transmission system and / or a data input system. The data transmission system can be designed in such a way that data is transmitted to the electronic data processing unit by means of a special PCI plug-in card or by means of a USB bus or by means of a Firewire interface. A data entry system can be a keyboard, for example.

According to a possible embodiment, the device is designed as a compact and mobile assembly. The device should be made as small as possible so that it can be easily carried by one person and possibly also used as a "stand-alone" unit. Since the demands on the electronic data processing unit are relatively low, in particular when pressure sensors are used exclusively in terms of their computing power, it is possible to design the necessary hardware components so small and to fit them to a person in such a way that they do not interfere with them Represent exercise, for example, a sporting movement. It is thus advantageously possible to integrate expert knowledge into a relatively small and easily transportable (mobile) unit in such a way that the determined posture data can be evaluated in real time and corresponding information can be transmitted quickly to the person.

Further advantageous embodiments of the invention result from the description.

The invention is explained in more detail in an exemplary embodiment with reference to an accompanying drawing. Show it:

FIG. 1 shows a schematic illustration of a device according to the invention;

Figure 2 is a block diagram of the inventive method and

Figure 3 shows a section of a flow chart of the training program.

FIG. 1 shows a schematic illustration of a person, generally designated 10, on whom a posture control is carried out by means of a device 12. The device 12 has a recording unit which serves to detect posture and which comprises two video cameras 14, a microphone 28, pressure sensors 16 and acceleration sensors 18. The recording unit is operatively connected to an electronic data processing unit 20 which has an expert database 22. The Electronic data processing unit 20 is used to evaluate the posture data of the person 10 transmitted by the recording unit. In this case, posture-relevant information is created using the expert database 22, which is transmitted to the person 10 by means of an information output unit that contains a monitor 24 and two loudspeakers 26. The loudspeakers 26 are connected to the data processing unit 20 (double arrows 38). The electronic data processing unit 20 has a data transmission system (double arrows 32) and a data input system 34. The data can thus be transmitted to the electronic data processing unit 20 by means of data carriers, by manual input and / or by online transmission (for example the Internet). Posture-relevant data can be transmitted to the electronic data processing unit 20 by means of the recording unit (video cameras 14, pressure sensors 16, acceleration sensors 18, microphone 28) by means of suitable data transmission lines (double arrows 34, 36) or also wirelessly. A plurality of optical marking elements 30 are attached to person 10 in order to create reliable data by means of video cameras 14.

The device 12 is used to determine data relating to a particular body position, body movement and body strain (posture) of the person 10 by means of the recording unit, to supply these posture-relevant data to the expert database 22 of the electronic data processing unit 20 and from this easily and generally understandable information or Instructions for extra here and to the person performing a sporting exercise 10, for example. A closed control loop is thus formed in the form of a "bio-feedback process". It is advantageously possible, for example, to control movement sequences of person 10 in the form of a bio-feedback loop, to report dangerous situations and / or to give complex course and training instructions. This is done electronically with the help of a computer (electronic data processing unit 20), so that the use of a human teacher or expert can be dispensed with.

The electronic data processing unit 20 according to the exemplary embodiment can be a commercially available computer (PC with at least 200 MHz and at least 64 MB RAM). The optical data recorded by the video cameras 14 are digitized in the form of analog video data using so-called frame grabbers into 8-bit wide data and then instructed in the electronic data processing unit 20, in which the further evaluation takes place using suitable software. Alternatively, web cameras can be used that provide the image data directly in digitized form. The evaluation can also be carried out on fast 16-bit microcontrollers (not shown). Depending on requirements, additional posture data of the body of the person 10 can additionally or alternatively be recorded by means of pressure sensors 16 and / or acceleration sensors 18. Standard sensors can be used here. The determined by the sensors Posture data are recorded and stored using a microcontroller system. In the case of acceleration sensors, the data ascertained are relatively complex, so that the actual data analysis takes place by means of a neural network which has been appropriately trained on a third-party computer. The weighting factors determined after training the network are then transferred to the microcontroller system.

For example, pressure distribution data are obtained by means of the pressure sensors 16, which are first standardized to the so-called body weight pressure by means of knowledge of the body weight and shoe size of the person 10 and are stored in the electronic data processing unit 20. EEProms or flash memory cards serve as memory. The capacity is large enough to store a possible running distance of 10 to 20 km, for example. The storage is used either for statistical purposes or for subsequent evaluation on a PC. The time course of the pressure distribution of the shoe sole is used to analyze the different gait phases (initial contact, stress response, middle class, terminal position, swing phase). From this data, various parameters are then derived (gait speed, step frequency, symmetry parameters, dynamics of the pressure center, assessment the rolling behavior). These secondary parameters form a multi-dimensional feature vector, by means of which a database in which the actual expert knowledge is stored is now queried. In such an expert database, both information and tion about dangerous situations as well as training instructions.

When optical marking elements 30 are used on the person 10, the markings recorded by the video cameras 14 are recognized by the electronic data processing unit 20 by means of a so-called threshold operation, and their coordinates are calculated. By comparing coordinates of the same points in each case in two corresponding recording images of the two simultaneously recording video cameras 14, the third dimension can be calculated with a known distance between the video cameras 14. Alternatively, the coincidence method according to Rolf D. Henkel, which is described in the publication "Synchronization, Coherence-Detection and Three-dimensional Vision, Perception, Neural Dynamics & Spiking Neurons, 2000", can be used. The respective position of person 10 is calculated from knowledge of the spatial position of these markings. These posture-relevant data are standardized and compared with the expert database 22. The position of the person 10 per learning unit is exactly stored there. Depending on the deviation, appropriate information or instructions are now created for the exercising person 10. The information or instructions are transmitted to the person 10 by means of the information output unit (monitor 24 and / or loudspeaker 26). The information or instructions are preferably transmitted by means of natural language voice output. For example, person 10 receives correction instructions as long as until a corresponding correct exercise position is exactly taken. Then you can proceed to the next part of the course.

The structure of the expert database 22 is explained on the basis of a stereo-optical determination of the body posture of the person 10 by means of video cameras 14 and on the other hand also for the evaluation and analysis of the pressure distribution via a running shoe (pressure sensors 16).

From the stereo-optical image data, the angles that the different markings made on the body form with one another are calculated and stored. The markings are placed in such a way that the resulting angles correspond to the position of joints (e.g. knees, hips). A body position is clearly defined by a combination of these different marking angles. There is an entry in a table for each of these different marking angles (for example left arm, right arm, left leg, right leg, head, hip, etc.). The current marking angles are compared with the corresponding specifications in the table. If all angles are within a specified tolerance, the line number defined in the same line is continued. Certain criteria such as exceeding or falling short of time lead to the continuation of an alternative line number. If the measured marking angles deviate from the specification in the table, text instructions are generated using line and marking-specific text identifiers. Each row in this table thus corresponds to a specification for a specific posture. In addition to these specifications, there are additional entries in the different columns for the individual marking angles. The entire entries can be roughly divided into four different groups: one or more so-called default groups, one or more extreme value groups, a time limit group and a branching group (see Table 1). If necessary, additional groups can be defined.

Table 1 (Structure of a data line)

Each default group consists of a valid flag, a lower and upper limit for the angle, two different text identifiers, a priority value and a value for transitions (see Table 2). The valid flag decides whether the whole group is active and is thus observed or not (don't care state). The upper and lower limit defines the tolerance range within which an angle is recognized as matching. If the measured marking angle lies outside this tolerance range, a text output is generated. The text identifier defines the type of instruction to be generated (for example "lifting" or "lowering", "stretching", "bending" etc.). The text identifier exists twice, depending on whether the angle is too large or too small. If the measured angles of several default groups at the same time An associated priority value decides which message is to be output outside of its tolerance ranges and thus multiple text outputs are generated. The number of transitions defines how often the valid vault area can be left (the practitioner wiggles back and forth) before the termination criterion is reached.

The extreme value group is similar to the default group with the difference that the entries in it are used to warn of positions that are hazardous to health. In the time limit group there is an entry for the average time it should take to take the posture. Exceeding this value fulfills the termination criterion. A second time specifies how long the posture should be maintained. The branch group regulates in which line to continue next. There is one line number for the case that all conditions are met and one for the termination criterion. It can be either an absolute jump or a subroutine call. Accordingly, there is an entry that corresponds to a return statement.

Table 2 (Structure of a default or extreme value group):

The ability to define subroutines offers various options for the actual course of the tutorial. On the one hand, the entire program can be stored in a single large table and is called up exactly once by the computer program; the overall sequence is in the table. On the other hand, the learning material can also be coded in various smaller tables that are called up several times by the computer program. This is particularly useful for the formation of loops and the like. The overall process in this case is in the order in which the tables are called from the main program of the computer.

The generation of the text instructions depends on the one hand on the text identifier of the default group and on the other hand on the deviations of the measured marking angles from the target value. The text identifier identifies the text to be output more or less directly. There are two identifiers because the instruction text depends on whether the measured value is above or below the tolerance range. If the tolerance range is not reached in several default groups, in the simplest case the priority field decides whose instructions are to be issued. In the event of large or small deviations of the setpoint from the actual value, the output text is modified accordingly ("very", "a little", etc.). In order to make the output text more natural, the text can be generated using fuzzy logic with appropriate rules from several default groups with their text ID, priorities and deviations from the target value. (if text identifier i is up and text identifier ₂ is forward and priority i is equal, then text is erected immediately). If the termination criterion is met, for example due to the time being exceeded, the program continues at a different and perhaps "easier" point in the sequence program.

A movement can be defined by exactly one or more different lines. The more different lines are assigned per movement sequence, the more precise instructions can be given. For example, the starting position of an exercise could be defined by several lines. First, for example, only the trunk position would be entered per line, the rest would not be in the care state. The arm position would be entered in a second line in addition to the trunk position. The leg position could be added in the next line, this line would have entries for the three default groups trunk, arm and leg position at the same time. Leaving the tolerance range within the body default group would automatically generate a text according to the text identifier of this line. If the arm area is also left at the same time, in the simplest case the priority value decides which message is generated. If, for example, the leg position cannot be reached in the third step, the termination criterion is generated after the time limit has expired and the program is continued at another point. Alternatively, the program can be repeated at the same point, but with a different difficulty Level to be continued. This level can modify all default values, such as the tolerance range. If the skill of the practitioner increases after a certain time, his level also rises and therefore the requirements must be adhered to more precisely. The principle of the difficulty level simplifies and reduces the tables considerably.

The principle of the program is based on the fact that the (one-off) coding of the learning program itself is relatively complex, but the execution is all the easier and therefore faster. This is achieved in that each necessary sub-process is represented by its own column entry in the table. The resulting table is very extensive, but can be easily evaluated due to the fixed format. Each entry in a column controls an action, comparable to a micro program on a computer. A disadvantage of this method is the relatively large memory requirement of the table. Assuming that 10 marking angles are monitored, this results in approximately 200 bytes / line and thus approximately 200 kB for 1,000 lines. However, this is not a problem with the memory sizes available today.

The type of coding of the specifications using the table presented above has the additional advantage that the evaluation can be easily parallelized. Each default group monitors the corresponding inputs completely independently of the others and generates the corresponding instructions. The individual results are only in a last step summarized to an overall result, in the simplest case about the existing priorities. Due to the simple structure, the evaluation - if necessary - could be carried out entirely by means of its own hardware logic (PAL modules). It would be easier to use a separate microcontroller (Pic controller) for each group of specifications, which forwards the results to a central controller.

When used for pressure sensors, the angular tolerance ranges are replaced by other parameters, for example pressure values. The tutorial consists of much fewer different lines, so that the available EEProm area is sufficient. If a more extensive learning program is desired, one or more additional microcontrollers can be used instead of a simple memory expansion because of the simple parallelization of the evaluation, as discussed above. In addition to the memory management, this greatly increases the evaluation speed.

The device 12 thus enables a reliable, quick and learning-friendly posture control on a person, for example when performing a sporting exercise or also when learning a musical instrument, without necessarily having to resort to a human expert.

The structure of a system for evaluating and analyzing the pressure distribution over a running shoe is described below. FIG. 2 shows the sequence of the method according to the invention in the form of a Block diagram shown. At the entry level, the data from various sensor modules are synchronized and combined. In the sensor modules 1 and 2, print data are determined from N data channels each. In addition, the data is stored and preprocessed in sensor modules 1 and 2. The preprocessing affects all processes in which only the information of a single sensor module is required. The data is then forwarded to a master module, where further processing takes place. The data channels are synchronized in the block Synch.

At the filter level, the data is filtered and standardized over time and space. The input data FIFO contains the raw data. The filter data FIFO contains the data after the filter operation.

The block reference pattern stands for various reference patterns that are required for processing in the object level. The reference patterns are used to determine correlation factors in comparison with the data in the filter level.

In the object level, all objects are defined that are accessible from the expert database. Every object has a set of conditions, a derived truth value, properties and methods. The conditions relate either to the data at the filter level or to the properties and methods of objects from which they are derived. The objects of the first level (object 11, object 12, ..., object In) directly access the filter level. The derived objects objects (object 21, object 22, ..., object 2n) access the data of the filter level indirectly via the objects of the first stage, from which they are derived.

The actual training program runs in the expert system. The training program can be implemented in the form of hardware (own microcontroller, programmable logic) or as software. The expert database contains the totality of all sub-steps of the entire training program in the form of individual program lines. Several object conditions are defined in each line with the associated branching to other program lines. When the condition is met, the corresponding branching is carried out. The interpreter with program counter, stack and registers controls the flow of the expert database.

At the output level, the output is generated in the form of natural language or tones and melodies. The instructions for this come from the expert database. In the case of a speech output, the instructions are generated from stored speech sequences. When a melody is output, tones and melodies are generated using MIDI commands.

The objects are links between the actual sensor data and the expert database and represent an abstraction layer. This makes the actual expert database independent of the hardware used and greatly simplifies programming and any changes. An object has its own probability value between zero and one, which depends on the one hand on the input data itself and on the other hand on the degree of fulfillment of conditions defined in the object. In addition, an object has properties and methods for changing these properties. An object can be derived from another object and thus inherits both its probability value and its properties and its reference to the sensor data record. The own probability value then results from a combination of the fulfillment of the own conditions and the inherited probability value. First-order objects can directly access the original or filtered sensor data. Objects derived therefrom can access indirectly via the data reference of the objects of the first order from which they were derived.

The method is explained below using a simple example. An object “pecodicity” is defined in which sensor pressure data are examined to determine whether rhythmic fluctuations occur. For this purpose, the distance between two extreme values is determined in the low-pass-filled pressure data, which are available in the FIFO. If the distance meets certain minimum / maximum requirements and has a certain fluctuation range, the probability value is set accordingly by a fuzzy rule set. The requirements are part of the object "Peπodizitat".

Another "Step" object is defined. This object refers once to the object periodicity of the left pressure curve PL and other on the other hand on the periodicity of the right pressure curve PR. The conditions defined in the "Step" object require that both PL and PR must be true. In addition, the PL and PR values must be in opposite phases. If the condition is met, the "Step" object is assigned the corresponding fuzzy probability value. As a property, it is assigned the numerical value of the phase shift. An object "bouncing" would look very similar, only that the conditions would provide an in-phase pressure curve.

Derived from the "Step" object, the "Walking" and "Running" objects can now be derived. The probability value of these modules depends on the phase shift property of the "Step" object. The exact conditions are specified in the objects "walking" or "running". A property of the "walking" object could, for example, be the walking time that can be queried in the object from the expert database. Differences between old and new values could also be provided as properties.

A module "overpronation" could be derived from the object "step". This module would use the reference to the print data from the object "step" to check to what extent the pressure values in the data identified by the object "step" correspond to the required conditions and set its probability value and its properties accordingly. Statistical objects can also be defined by other objects Add the measured times to a total time or to energy consumption.

In Figure 3, the sequence of the method is shown as an excerpt. The procedure in line 1 requires that 10 minutes be left. Running, bouncing or a strong overpronation would trigger a corresponding correction instruction. After the 10 minutes of walking, it continues in line 2. Row 4 would continue after 12 minutes at the latest.

The line number serves as a reference for the program run. The output message (message) is a clear reference for the voice or sound output to be generated. Alternatively, a flag can be set instead of a voice output, which can be queried in the Conditions column. In the Conditions column, all conditions are recorded to which you want to react. The conditions can refer to objects themselves or their properties. On the other hand, standard objects such as the Z time, which contains the dwell time per line, can be queried, which are automatically set each time a line is processed. The Priority column specifies which condition should be addressed first if several conditions are met at the same time.

The Object manipulation column indicates whether an object should be changed using its methods. The PCounter column specifies how the training program counter should fulfill the the condition is to be changed. There are simple jumps (Goto), subroutine calls (Call, Return) and loop instructions (Do, While, For, Next, Loop). There are also unconditional jump instructions that combine different, recurring queries into one block. The Max. # Column is to be seen in connection with the program counter control. With call commands, it shows the maximum number of times the subroutine can be called.

The reference points are recorded by means of several video cameras, which are mounted at a certain angle and distance from one another. The distance between the points and the cameras is determined from the offset of corresponding points in related image rows. A coincidence procedure is used for this.

The position of the reference points directly reflects the attitude of a person in the room. The relationships of these points to each other can therefore directly address the expert knowledge in the form of the training program. In the case of modules in which sole pressure sensors (for example jogging module, ski module) and / or other sensors (structure-borne noise, acceleration sensors, inclinometers) are used, the posture or movement must be determined indirectly using the data recorded by the sensors getting closed.

For example, the recognition or differentiation of the movements "walking" and "running" is only possible through the temporal pressure curve of both sons. to differentiate len. In skiing, for example, a parallel turn is defined by the time sequence of a complex pressure pattern on the outer and inner skis. An additional preprocessing stage is required to carry out this type of recognition. Indirect movement data is used to infer the actual movement and posture.

The preprocessing stage is preferably carried out in a two-stage process for deriving movement data from indirect movement data. A secondary data record is derived from a primary data record. The movement or behavior is then concluded from the two data sets together. For this purpose, the temporal data stream of the two pressure-sensitive soles or the other sensors is temporarily stored in a FIFO buffer and filtered. Two methods are used for filtering. First, the filtering using a one- or multi-dimensional matrix (kernel with the weighting factors), the so-called convolution. With this method, all common filter operations such as high pass, low pass, tophet, mean, median, max, ming, sobel, roberts etc. can be carried out with comparable effort in the time domain. Secondly, the data are subjected to FFT or wavelet filtering. These operations take place in the frequency domain and provide more precise information about the frequency spectrum in general (FFT) or per space / time unit (wavelet). In addition to filtering, other secondary data, such as the course of the center of gravity (COG) per sole and its deviation from a reference straight line, the percentage pressure distributions etc. are determined. The derived secondary data is used together with the primary raw data stored in its own FIFO buffer and additional antropomorphic data to identify posture or movement. The primary and secondary data must be temporarily stored, since some movement patterns are only defined by the chronological sequence of individual movement components. Since there is a wide variation in the movement sequence from person to person, the analysis of the primary, secondary and antropomorphic data is expediently carried out using fuzzy fuzzy logic in conjunction with a rule database. As a result of this preprocessing level, movement and posture information, such as standing, walking, running, jumping, climbing stairs, etc., but also statements about the dynamics of the movement, such as powerful, sluggish or tired, are available over time ,

With the ski module, this not only enables simple statements, such as supine position, template, one-sided, double-sided, full-surface loading, edge loading, but also complex statements, such as the beginning of a curve, the end of a curve, plow, parallel turn, short swing, carving, etc. is not possible with simple threshold systems.

In a comparable procedure, harmful movement components are based on the analysis described identified by complex print patterns over time. This analysis uses additional internally recorded statistical data on the duration and frequency of certain postures and movements. Whether a movement component is harmful also depends on how often it is carried out in which period. These statements are not available in simple threshold systems.

The movement sequences recognized in the preprocessing stage, including the harmful movement sequences, can be made audible by means of a tone to enable a bio-feedback process. Instead of a simple tone, a MIDI-compatible sound module is preferably used.

The statements about the behavior determined in the preprocessing are now fed to an expert system in the form of training instructions, similar to the video module. This program now makes it possible to react to the person's behavior, to warn and to carry out a complete training program. Training program can also be understood as a program for health or fitness check or for determining the accuracy of fit of footwear etc. What is generally meant is a sequence of instructions to the user, the user's reaction being assessed and the further course of action being specified.

The ski module also uses sensors that are suitable for recognizing the distance and the angle of the skis from one another. Small portable generators are used for the power supply sets that use either the movement of the exerciser / sports device or the temperature difference between the exerciser / sports device and the environment to generate electricity.

The pressure sensor unit (sensor and receiving unit) can also be designed as a complete stocking. The data processing unit and information output unit can be incorporated into a piece of clothing, for example a jacket.

Claims

Expectations

1. Procedure for posture and / or movement control of at least one person using a training program, with the following procedural steps:

a) posture-relevant data are recorded with the aid of several sensor devices;

b) the data recorded by the sensor devices are put together and in particular synchronized;

c) the composite and synchronized data are filtered and in particular standardized and exempted;

d) the filtered and standardized data are compared with reference data in order to determine correlation factors and / or are examined for the occurrence of characteristic features, such as extreme values of angle values or the exceeding or falling below threshold values;

e) using the correlation factors and / or the characteristic features that occurred

Defines objects, each comprising a set of conditions and a probability value derived therefrom; f) the objects are processed in an expert database and, depending on the fulfillment of one or more conditions and the current status of the training program, messages are output and the further course of the training program is controlled.

2. The method according to claim 1, characterized in that the messages are output to the person while the relevant training unit is being executed, in particular immediately after the associated conditions have been met, in particular in real time.

3. The method according to any one of the preceding claims, characterized in that the training program takes into account the person's reaction to the messages output.

4. The method according to any one of the preceding claims, characterized in that the expert database comprises a learning program which has a plurality of lines and the sequence of which ensures by a counter which designates the line on which the learning program is currently located.

5. The method according to claim 4, characterized in that each line of the learning program comprises an instruction block, a control block and a branch block.

6. The method according to any one of the preceding claims, characterized in that the posture data by means of an optical and / or acoustic Auf- acquisition unit (14, 28) and / or by means of sensors (16, 18).

7. The method according to any one of the preceding claims, characterized in that the determined posture data are analyzed in the data processing unit (20) for the subsequent derivation of parameters that are compared with data in the expert database (22) to create at least one posture-relevant information for the person (10).

8. The method according to any one of the preceding claims, characterized in that a posture response to information is used as a criterion for personal assessment and for creating an adapted information creation.

9. Device for carrying out the method according to one of claims 1 to 8, with at least one recording unit serving for posture detection, characterized in that an expert database (22) and posture data evaluation (20) is provided electronic data processing unit is provided with the sensor devices and / or the recording unit (14, 16, 18, 28) and is operatively connected to an information output unit (24, 26) serving to transmit at least one posture-relevant information to the person (10).

10. The device according to claim 9, characterized in that in addition an acoustic recording Unit (28) is provided, which is operatively connected to the data processing unit (20) for sound data evaluation.

11. The device according to one of claims 9 or 10, characterized in that the receiving unit

(14, 16, 18, 28) has at least one sensor (16, 18) for recording posture-relevant data.

12. Device according to one of claims 9 to 11, characterized in that the sensor (16, 18) is designed as a pressure sensor (16) or as an acceleration sensor (18).

13. Device according to one of claims 9 to 12, characterized in that the recording unit (14, 16, 18, 28) has at least one video camera (14).

14. Device according to one of claims 9 to 13, characterized in that at least one optical marking element (30) is provided for attachment to the person (10).

15. Device according to one of claims 9 to 14, characterized in that the information output unit (24, 26) is designed as a transmitter for optical and / or acoustic and / or tactile signals.

16. The device according to one of claims 9 to 15, characterized in that the acoustic signals are speech signals.

17. Device according to one of claims 9 to 16, characterized in that the data processing unit (20) has a data transmission system (32) and / or a data input system (34).

18. Device according to one of claims 9 to 17, characterized in that it is designed as a compact and mobile assembly.