CN103079289A - Wireless multi-person motion data collector and collection method - Google Patents

Abstract

The invention discloses a wireless multi-person motion data collector and a collection method. The wireless multi-person motion data collector comprises tens of data sensing nodes, six data relay nodes and a data convergence node, wherein the data sensing nodes are adhered to each motion key position of the human body; each human body motion data is periodically collected by a high-precision acceleration sensor, a magnetic sensor and a gyroscope; compressed sensing data is wirelessly transmitted to the data relay nodes; the data relay nodes are in charge of integrating data and transmitting the data to the data convergence node; the data convergence node is in wired connection with six data relay nodes; and the data is periodically uploaded to a computer. According to the invention, the motion data of 2-4 people can be simultaneously and wirelessly collected, and the invention has the characteristics of high precision, high data rate and low power consumption and is widely suitable for occasions, such as virtual man-machine interaction, virtual games and limb medical rehabilitation.

Description

A kind of wireless many people motion data collection element and acquisition method

Technical field

The present invention relates to data acquisition and wireless sensor network technology field, relate in particular to a kind of wireless many people exercise data centralized procurement storage.

Background technology

Exercise data acquisition is the key component of action capture technique, by the action capture technique, computer is appreciated that human action, then the user just can send instruction, convey a message etc. to computer by mode such as figure, orientation, gesture and expressions, except field of human-computer interaction, motion-capturedly can also be applied in the fields such as motion analysis, model based coding, virtual reality, cartoon making, intelligent monitor system, game making, the motion capture technology that is applied to human body has broad application prospects and huge commercial value.

Since the eighties in 20th century, the U.S. has several laboratories to carry out successively the research of human movement capture system, such as Biomechanics laboratory, Simon Frase university, the Massachusetts Institute of Technology etc.After this, along with increasing researcher and developer's discovery the value of movement capturing technology, this technology also little by little moves towards practical from Journal of Sex Research on probation.At present in developed countries such as America and Europes, the high request that progressively improves in order to adapt to cartoon making, movement capturing technology has entered practical stage, many manufacturers its commercial motion capture device that released one after another is arranged, such as MotionAnalysis, MAC, Polhemus, FilmBox, Sega Interactive, X-Ist etc., the business-like application of success has the biomechanics Research such as athletic training, virtual reality, game interactive, human body etc. at present.

Existing commercialization motion capture system mainly is divided into five classes from operation principle: microsensor MEMS formula, electromagnetic type, optical profile type, mechanical electric dynamic formula, acoustics formula.Generally can estimate distinct device from the following aspects: data precision; Can catch Moving Objects number what; Can catch the space range size; Real-time; Antijamming capability; Degree easy to use; Cost etc.

Present existing wireless motion data collection element, the wireless data sensing node quantity that mainly exists support is few, and the Refresh Data rate is low, simultaneously the problem such as many people of high speed acquisition exercise data.

Summary of the invention

The object of the invention is for the deficiencies in the prior art, and a kind of wireless many people exercise data centralized procurement storage and acquisition method are provided.

The objective of the invention is to be achieved through the following technical solutions: a kind of wireless many people motion data collection element, it is characterized in that, it is mainly by dozens of data sensor node, six data via nodes and a data aggregation node form, three kinds of nodes form radio sensing network based on sub-clustering in order to accurately, gather at high speed every human motion sensing data.Wherein, dozens of data sensor node is attached to each motion key position of human body, after the data compression that periodically gathers, be sent to the relay station that connects by wireless network, convergence node and six data via nodes are by SPI interface wired connection, tdm communication, sensing data gathers to computer by USB the most at last.

A kind of above-mentioned many people exercise data acquisition method based on the clustering wireless sensor network network, it is characterized in that the method comprises Time Synchronization Mechanism, the pipeline data polymerization, multichannel room avoidance mechanism and local attitude quaternion filtering method of estimation may further comprise the steps:

(1) convergence node and relay station initialization.

(2) radio sensing network is set up in convergence node and relay station networking, waits for that the data sensor node networks.

(3) data sensor node initializing.

(4) data sensor node monitor channel, application networks.

(5) data sensor node periodicity pick-up transducers data, and process this secondary data of compression.

(6) data of data sensor node after according to collection period encapsulation compression, and enter low-power consumption resting state waiting timer it is waken up.

(7) the data sensor node is according to the collection period timing wake-up, and sends data to the specific data via node.

(8) convergence node receive data regeneration node data, and be sent to computer by USB.

(9) DTDs, repeating step 5-8 realizes the periodicity collection of motion-sensing data.

The invention has the beneficial effects as follows:

1. adopt this kind wireless human body motion data collection element, can arrive simultaneously the refreshing frequency collection 2-4 people exercise data of 120Hz with 30Hz, and accurately data are sent to terminal by radio sensing network.

2. adopt Wearable data sensor node, utilization is based on three kinds of dissimilar sensors and the small-sized MCU+RF module of MEMS MEMS, reduce as far as possible the pcb board area, and appropriate design is dressed shell, not only guarantee the comfortableness of dressing, and effectively guaranteed the high performance requirements of node.

3. adopt the compensating filter structure based on gradient descent algorithm, go out to describe the quaternary numerical value of attitude matrix at data sensor node local computing, be less than 2 ° RMS precision, effectively reduced volume of transmitted data, save a large amount of wireless bandwidths, well solved the conventional motion data acquisition unit and carried data sensor number of nodes and band-limited contradiction.

4. adopt the pipeline data polymerization, comprehensive next as for same node difference data constantly, by merging homogeneous data and removal redundancy, less volume of transmitted data and time slot synchronization overhead further reach the purpose of saving bandwidth and energy consumption.

5. adopt the clock cycle to adjust to guarantee the strict slot synchronization of sensing network, reach the purpose that network is lower than 1% packet loss, guaranteed the stability of wireless system.

6. adopt multichannel room back off algorithm MARB, efficiently solve the networking race problem that can only allow node networking characteristics to cause because of the sensing network list networking cycle, make the sensing network maximizing efficiency, reach the purpose of quickly networking and the quick fault recovery of collector.

Description of drawings

Fig. 1 is system applies schematic diagram of the present invention;

Fig. 2 is data sensor node hardware block diagram;

Fig. 3 is data sensor node M CU+RF module principle figure;

Fig. 4 is the conditioning of data sensor node power and detection module schematic diagram;

Fig. 5 is data sensor node acceleration sensing module principle figure;

Fig. 6 is data sensor node M EMS numeral output gyroscope modules schematic diagram;

Fig. 7 is data sensor node magnetic induction sensor module principle figure;

Fig. 8 is convergence node and relay station hardware block diagram;

Fig. 9 is convergence node ARM module principle figure;

Figure 10 is convergence node USB module principle figure;

Figure 11 is relay station DSP module and radio-frequency module schematic diagram;

Figure 12 is convergence node and relay station power module schematic diagram;

Figure 13 is multichannel room back off algorithm MARB schematic diagram

Figure 14 is the compensating filter structural representation based on gradient descent algorithm;

Figure 15 is pipeline data polymerization schematic diagram;

Figure 16 is radio sensing network time synchronized time slot schematic diagram;

Figure 17 is data sensor node software flow chart;

Figure 18 is the relay station software flow pattern;

Figure 19 is convergence node software block diagram.

Embodiment

Appropriate design of the present invention realizes three kinds of node software and hardwares, adopt a kind of radio sensing network based on sub-clustering, realized a kind of data acquisition unit and acquisition method that can carry out at 30Hz wireless real-time tracking in the 120Hz data transfer rate refresh rate scope to 2-4 people's exercise data, it is high that it has the Refresh Data rate, it is many to carry the data sensor number of nodes, network packet loss rate is low, stable advantages of higher.

As shown in Figure 1, that this collector is used schematic diagram, this wireless many people motion data collection element comprises the data sensor node, relay station and convergence node, wherein the data sensor node is connected with relay station by the wireless network based on the IEEE802.15.4 standard, relay station is connected by the SPI interface with the convergence node, and the mutual co-ordination of three jointly forms a radio sensing network based on sub-clustering and gathers the transmission human body movement data to computer.

The below explains to the present invention from the realization of three kinds of nodes and the realization of acquisition method respectively.

One, data sensor node, relay station and convergence node hardware are realized

The Various Functions of three types node, power supply plan and energy consumption are different among the present invention, need to adopt different implementation methods.Sensor node adopts lithium battery power supply, and the energy content of battery is limited, and charging or replacing battery are unfavorable for the smooth and easy work of system continually, so the low power dissipation design that sensor node need to adopt; That relay station and convergence node need to have is stronger, faster data processing speed, so hardware designs will guarantee the reliable and stable of circuit, reduces the burden of software.

1, data sensor node hardware is realized

The function that the data sensor node need to be realized comprises collection, radio communication and the electric quantity monitoring of 3 kinds of dissimilar sensing datas.Its hardware system block diagram as shown in Figure 2.As can be seen from Figure, sensor node mainly comprises: MCU+ wireless radio frequency modules, acceleration sensor module, magnetic induction sensor module, MEMS numeral output gyroscope modules and power supply conditioning and monitoring modular.The MCU+ wireless radio frequency modules is connected with plate and carries antenna, and MCU passes through I ²The C bus links to each other with MEMS numeral output gyroscope modules, magnetic induction sensor module, links to each other with acceleration sensor module by the SPI interface.The data sensor node produces the supply power voltage that each hardware module needs by the power supply of 3.3V/250mAH chargeable lithium cell through the power supply conditioning module, the electric weight of power supply monitoring module monitors battery, and it is by common IO mouth Simulation with I ²The C bus protocol is connected with MCU.

As shown in Figure 3, be MCU+ radio frequency modular circuit schematic diagram, comprise that mainly MCU chip U1, crystal oscillator X101, Ba Lun T1, plate carry antenna ANT and some electric capacity, resistance and inductance.Wherein U1 can select the MC13213 chip of Freescale company, and its chip itself includes the radio frequency transceiving module based on 2.4GHZ, but is not limited to this.MCU chip U1 input pin 14,15 and 50,51 links to each other with power supply VCC by pull-up resistor R9, R8 and R15, R14 respectively, and power supply input pin 6,22,23,45 is respectively by shunt capacitance C3, C5, C4, C9 ground connection.MCU internal radio frequency module crystal oscillator pin 27,28 links to each other with crystal oscillator X101, and by capacitor C 7, C8 ground connection, power supply input pin 29,30,33 connects power supply VCC, and by capacitor C 1 ground connection, pin 32 connects inductance FB1, the inductance other end also links to each other with capacitor C 1, and pin 31 is by capacitor C 6 ground connection. Rf data pin 35,36 links to each other with 3,4 pin of Ba Lun T1 by inductance L 1, L2 respectively, the direct flying capcitor C11 of 3,4 pin of Ba Lun T1, and the balanced pin 34 of radio frequency level is connected with Ba Lun 2 pin, and by capacitor C 12 ground connection.Plate carries antenna ANT and is connected with 1 pin of Ba Lun T1 by capacitor C 13, and by inductance L 3 ground connection.The 5 pin ground connection of Ba Lun T1,6 pin are unsettled.

As shown in Figure 5, be the circuit theory diagrams of acceleration sensor module, comprise that mainly acceleration sensor chip U2, electric capacity are some.Wherein U2 can select 3 axle acceleration sensor MMA7455L of Freescale company, but is not limited to this.Acceleration sensor chip U2 power supply 1,6 pin connect power supply VCC, and by capacitor C 34, C35 and capacitor C 36, C37 ground connection, pin 2,3,4,5,10,11 equal ground connection.Sensor data interface pin 12,13,14 links to each other with MCU pin 17,18,19 respectively, and Data Control pin 7,8,9 links to each other with MCU pin 16,13,4 respectively.

As shown in Figure 6, be the circuit theory diagrams of MEMS numeral output gyroscope modules, comprise that mainly gyroscope chip U3, electric capacity are some.Wherein U3 can select 3 axle gyroscope ITG-3200 of InvenSense company, but is not limited to this.Gyro sensor chip U3 power supply 8,13 pin connect power supply VCC, and by capacitor C 33, C31 ground connection, pin 10,20 is respectively by capacitor C 32, C30 ground connection, pin 9,18 direct ground connection.Sensor data interface pin 23,24 links to each other with MCU pin 15,14 respectively, and Data Control pin 12 links to each other with MCU pin 3.

As shown in Figure 7, be the circuit theory diagrams of magnetic induction sensor module, comprise that mainly magnetic induction sensor chip, electric capacity are some.Wherein U4 can select 3 axial magnetic sensor MMC3120MR of MEMSIC company, but is not limited to this.Magnetic induction sensor chip U4 power supply 5,6 pin connect power supply VCC, and by capacitor C 38, C39 ground connection, pin 3 is by capacitor C 40 ground connection, and pin 1,2 connects capacitor C 41 two ends.Sensor data interface pin 7,8 links to each other with MCU pin 14,15 respectively, shares I with the gyroscope chip ²The C data-interface.

As shown in Figure 4, be the conditioning of data sensor node power and monitoring modular, mainly comprise lithium cell charging chip U7, lithium battery chip monitoring U6, voltage voltage stabilizing chip U5, toggle switch S1, lithium battery BT0, USB connector P1, power connector P2, power supply indication LED lamp and some electric capacity, resistance.Wherein U7 can select the MAX1555 chip of MAXIM company, but is not limited to this, and U6 can select the BQ27200 chip of TI company, but is not limited to this, and U5 can select the SPX3819 chip of SIPEX company, but is not limited to this.Voltage-output 5 pin of voltage voltage stabilizing chip U5 connect VCC, and by capacitor C 16 ground connection, voltage input pin 1 and chip enable pin 3 connect cell voltage end V_BAT, connect simultaneously toggle switch S1 one end, and by capacitor C 14 ground connection, 4 pin are by capacitor C 15 ground connection, pin 2 ground connection.Lithium cell charging chip U7 power supply input pin 1,4 connects respectively power supply V_USB, V_DC, and by capacitor C 17, C18 ground connection, the V_USB end connects LED_USB lamp positive terminal simultaneously, and link to each other with 3 pin of U7 by resistance R 5, the V_DC end connects LED_DC lamp positive terminal, and links to each other with 3 pin of U7 by resistance R 6.The U7 power supply is exported 5 pin and is linked to each other with lithium battery anode, and by capacitor C 19 ground connection.Lithium battery chip monitoring U6 power supply is inputted 2 pin and is linked to each other with lithium battery anode through resistance R 1, and connects simulation ground by capacitor C 20, and voltage detecting pin 6 links to each other with lithium battery anode through resistance R 2, and connects simulation ground, data I by capacitor C 22 ²C interface pin 4,5 links to each other respectively at MCU pin 50,51, and pin 1 connects simulation ground through capacitor C 21, and pin 3,11 connects simulation ground.Current detecting pin 7,8 connects simulation ground by capacitor C 24, C23 respectively, and links to each other with modulus resistance R s by resistance R 4, R3, resistance R s connecting analog ground and digitally.The cathode of lithium battery end links to each other with simulation ground.USB power supply or DC power supply can be provided by connector P1, P2 respectively.

2, relay station and convergence node hardware are realized

We adopt wired mode to communicate by letter with the convergence node by the design data via node, large for the collector data transfer rate, real-time characteristics, we adopt SPI connection data via node and convergence node, relay station and convergence node are integrated on the circuit board, by same power module power supply, hardware block diagram as shown in Figure 8.

The convergence node is as the interface of whole data acquisition system and computer, its task is by 6 data via nodes of SPI interface management, and by USB interface forwarding computer command and transmission human body movement data, it mainly comprises two modules: arm processor module and usb interface module.Collector has 6 data via nodes, and each relay station all can be managed and is no more than 8 wireless data sensing nodes, and it is responsible for processing and the forwarding of motion-sensing data, and it mainly comprises two modules: dsp processor module and radio receiving transmitting module.Dsp processor is connected with radio receiving transmitting module by the SPI mouth, and communicates by letter with arm processor by another SPI interface.Convergence node and relay station are powered by same power module.

As shown in Figure 9, be the circuit pinouts of convergence node arm processor module, mainly comprise arm processor U10.Arm processor can be selected the STM32F103VCT6 of ST company, but is not limited to this, and it has three different hardware SPI interfaces, by the GPIO mouth is extended to 6 SPI interfaces as the Slave chip select line, can support 6 data via nodes to communicate by letter simultaneously.

As shown in figure 10, be the circuit theory diagrams of convergence node usb interface module, mainly comprise USB connector P_USB, triode Q1 and some electric capacity, resistance and inductance.2 pin of triode Q1 connect the 3.3V power supply, and link to each other with 1 pin by resistance R 101, and 1 pin of triode Q1 links to each other with arm processor 35 pin, and 3 pin of triode Q1 link to each other with 3 pin of USB connector P_USB by resistance R 102.USB connector P_USB supply pin 1 pin connects inductance FB2 one end, the inductance FB2 other end connects the USB positive supply, and by resistance R 105 ground connection, data pin 2,3 pin link to each other with arm processor 70,71 pin by resistance R 104, R103 respectively, 4 pin are by inductance FB1 ground connection, and 5,6 pin link to each other and by resistance R 106, capacitor C 101 earths.

As shown in figure 11, be the circuit pinouts of relay station dsp processor module and radio receiving transmitting module, mainly comprise dsp processor Umid-1, radio receiving transmitting module JP1.Dsp processor can be selected the DSP MC56F8037 based on the 56800E kernel of Freescale company, but is not limited to this.Radio receiving transmitting module can be selected the 2.4GHz wireless radio frequency modules MC13201 of Freescale company, but is not limited to this.Collector comprises 6 relay station that ardware feature is consistent, and this sentences the individual data via node is that example is explained to its circuit pin.Dsp processor SPI1 mouth pin 2,42,33,32 pin link to each other with arm processor pin 29,30,31,32 pin respectively, and data terminal pin 21 pin link to each other with arm processor pin 3 pin.Dsp processor SPI2 mouth pin 38,37,44,45 pin link to each other with radio receiving transmitting module 13,10,12,14 pins respectively, and control pin 30,47,55,56 pin link to each other with radio receiving transmitting module 11,2,4,6 pins.

As shown in figure 12, the circuit theory diagrams of convergence node and relay station power module, mainly comprise voltage stabilizing chip U11, power connector P, power supply indicator LED_POWER, capacitor C 102, resistance R 107, wherein voltage stabilizing chip U11 can select the LM-1117 of ST company, but is not limited to this.2,3 pin of the power connector P ground connection that links to each other, 1 pin provides positive supply, link to each other with arm processor modular power source pin 6,73 pin, link to each other with dsp processor power pins 7 pin, and by capacitor C 102 ground connection, connect simultaneously power supply indicator LED_POWER positive terminal, power supply indicator LED_POWER negative pole end is by resistance R 107 ground connection.Voltage stabilizing chip U11 power supply input pin 3 pin connect 1 pin of power connector P, Voltage- output pin 2,4 output 3.3V voltages, 1 pin ground connection.

Two, based on many people exercise data acquisition method of clustering wireless sensor network network

The motion data collection element of the present invention's design, utilize the efficient characteristics of radio sensing network low energy, designed a kind of by consisting of the method that the clustering wireless sensor network network gathers the transmitting moving data between data sensor node, relay station and the convergence node, wherein utilize strict Time Synchronization Mechanism, the pipeline data polymerization, multichannel room avoidance mechanism and local attitude quaternion filtering method of estimation, so that continuously, fast, many people of wireless collection exercise data becomes possibility in real time, the method mainly realizes by following steps:

1, convergence node and relay station initialization.

Convergence node and relay station initialization comprise the initialization of arm processor, the initialization of dsp processor, radio-frequency module initialization.The initialization of arm processor and the initialization of dsp processor include bus frequency and set, the configuration of GPIO mouth, interrupt configuration, SPI interface configuration etc.; The radio-frequency module initialization comprises channel fix, clock frequency setting etc.

2, radio sensing network is set up in convergence node and relay station networking.

The convergence node is according to computer command setting network refreshing frequency, and periodically sends the networking beacon and communicate by letter with relay station by the SPI interface, carries out the ground floor networking of radio sensing network.Its detailed process is that the convergence node is to all repeat broadcast networking commands that is connecting, relay station is successively to convergence node log on address, the convergence node is for each relay distribution network address and the information such as wireless channel that relay station is used are set, so that relay station can begin the networking of the second layer and data sensor node smoothly, and can mutually not conflict.Relay station and convergence node networking process mainly comprise following substep:

(1) convergence node broadcasts networking commands.

(2) relay station log on address.

(3) information such as convergence node distribution network address, time slot position and configurating channel.

(4) the convergence node is confirmed, sets up radio sensing network.

(5) relay station periodic broadcast networking commands waits for that the data sensor node networks.

3, data sensor node initializing.

The data sensor node initializing comprises the initialization of MCU+ wireless radio frequency modules, acceleration sensor module initialization, magnetic induction sensor module initialization, the initialization of MEMS numeral output gyroscope modules.The initialization of MCU+ wireless radio frequency modules comprises the configuration of GPIO mouth, interrupt configuration, channel fix, clock frequency setting etc.; Three kinds of sensor die initialization blocks comprise setting state, data initialization etc.

4, data sensor node monitor channel, application networks.

Data sensor node and relay station form the second layer of this collector wireless sensing collection network.Relay station periodic broadcast networking commands, the data sensor node is selected suitable relay station, to its log on address.Data sensor node and relay station networking process mainly comprise following substep:

(1) relay station broadcast networking order.

(2) the data sensor node is searched for legal relay station according to back off algorithm, and to its log on address.

(3) information such as relay station distribution network address, time slot position and turnover rate.

(4) the data sensor node is confirmed the relay station response message, and successfully networking then enters next step, and is failed then carry out back off algorithm, repeating step (2)-(4).

(5) enter the data acquisition transmit stage.

If the networking collision problem occurs, the data sensor node will be carried out back off algorithm, and the time delay back off algorithm of node application networking competition specifically is expressed as follows:

What the communication physical layer between data sensor node and the relay station was used is the radio receiving transmitting module (IEEE802.15.4 wireless standard) of 2.4GHz ISM band.The room tabulation is the tabulation of the self-defining record time slot application situation of this collector systems, and it is created in the relay station, and when there being the data sensor node successfully to apply for networking, N is counted in the room of corresponding relay station _Rest-1, when the data sensor node off-line is arranged, N is counted in the room of corresponding relay station _Rest-1.Each data sensor node has its unique numbering (physical address Phy_addr) because will distribute unique time slot for each data sensor node, so each cycle at most only allow a node successfully to network.

As shown in figure 13, in a communication cycle, suppose to occur the competition of N data sensor node, compete simultaneously relay station 1 such as the first round, each data sensor node can take turns doing following operation so:

(a) change channel

Chanel=floor(N _chanel·Random()) (I)

Wherein Random ∈ [0,1), the maximum integer less than or equal to x, N are got in function f loor () expression _ChannelThe control channel number that expression is maximum, i.e. relay station number.

Behind formula (I) replacing channel, again network to corresponding relay station application, second taking turns of meaning as shown, to calculate the channel of replacing be 2 for tentation data sensing node 1 and 3 through types (I) herein, the channel that data sensor node 2 calculates replacing is 1.Data sensor node 2 is not because there is other node of competing simultaneously networking, then takes turns successfully at this that application networks, and the node that data sensor node 1 and 3 etc. continues to occur race condition then enters operation (b).

(b) time delay is kept out of the way in the room

Through operation (a) afterwards, under a refresh cycle, the channel that changes to does not have the data sensor node of conflict will successfully apply for networking, and the node that still conflicts, then time-delay is kept out of the way, and channel is made as first channel again.Its periodicity of keeping out of the way is made as

BackoffTime=floor(N _rest·Random()) (II)

N wherein _RestBe its relay station room number that receives.

Then through the data sensor node delayed when keeping out of the way will be after time delay be waken up request channel again, running into when competition can be again change successively channel or continues to keep out of the way to the order that operates (b) according to operation (a), until all data sensor nodes all network.

5, data sensor node periodicity pick-up transducers data, and process this secondary data of compression.

The data sensor node passes through I according to the frequency acquisition of relay station notice ²C bus and SPI interface gather three kinds of sensing datas of 9 axle 18Byte.In order to save wireless bandwidth, reach the purpose of supporting that many personal data are caught, namely increase entrained data sensor number of nodes, need to carry out the active data compression optimization to sensing data.This collector adopts local attitude quaternion filtering method of estimation to calculate the attitude quaternion of this image data at the data sensor node to sensing data, reaches the data compression purpose by transmitting 8Byte four element datas.

Transducer hypercomplex number account form has adopted the compensating filter structure based on gradient descent algorithm, has merged angular speed attitude method and vector observation.

(1), angular speed calculation attitude method

Relative x under the three-axis gyroscope output body axis system, y, the angular speed of z axle is designated as respectively w _x, w _yAnd w _z(unit is these three angular speed parameters: rad/s) at the inner vector that is designated as of three dimensions (body axis system) Be expressed as formula (1):

$\stackrel{\→}{w}_b=\left[\begin{array}{cccc}0& w_x& w_y& w_z\end{array}\right]---\left(1\right)$

Then geographic coordinate system relative to the hypercomplex number rate of change that body axis system rotates is

For:

$\stackrel{\·}{Q}_g^b=\frac{1}{2}Q_g^b\⊗\stackrel{\→}{w}_b---\left(2\right)$

Be located at constantly t, the attitude quaternion of the relative body axis system of geographic coordinate system is

And

Knowing under the initial condition, can be by the hypercomplex number rate of change Integration obtains.Their relation is suc as formula shown in (3) and the formula (4):

$\stackrel{\·}{Q}_{w,t}^g{}_b=\frac{1}{2}Q_{\mathrm{est},t-1}^g{}_b\⊗\stackrel{\→}{w}_t^b---\left(3\right)$

$\stackrel{\→}{Q}_{w,t}^g{}_b=\stackrel{\→}{Q}_{\mathrm{est},t-1}^g{}_b+\stackrel{\·}{Q}_{w,t}^g{}_b\mathrm{\Δt}---\left(4\right)$

In the formula,

Be the angular velocity vector of moment t, Δ t is sampling period or the refresh cycle of being called of sensing data,

For the attitude quaternion that obtains is estimated in a upper sampling instant.Subscript w represents that the computational methods of attitude quaternion are the angular speed calculation method.

(2), the vector observation is calculated attitude

A 3-axis acceleration sensor can be exported gravitational field and because the linear acceleration that the linear movement of node data sensing node causes of data sensor node under the body axis system.Equally, triaxial magnetic field sensor will be measured the magnetic field of the earth and when the size and Orientation in earth magnetic field.

Be located under the geographic coordinate system, magnetic field of the earth or gravitational field parameter vector that the data sensor node obtains are Under body axis system, the parameter vector that the data sensor node obtains is

Then in the free from error situation of ideal, can get:

$\stackrel{\→}{s}_b=\stackrel{\→}{Q}_g^b*\⊗\stackrel{\→}{d}_g\⊗\stackrel{\→}{Q}_g^b---\left(5\right)$

In the formula, Be the attitude quaternion that the relative geographic coordinate system of body axis system turns over, i.e. the attitude quaternion that turns over of data sensor node.

But, under the actual conditions, because there are error in the acceleration of gravity that records and the data of geomagnetic field intensity, so the equal sign in the formula (5) and being false.The definition error function is:

$f(\stackrel{\→}{Q}_g^b,\stackrel{\→}{d}_g,\stackrel{\→}{s}_b)=\stackrel{\→}{Q}_g^b*\⊗\stackrel{\→}{d}_g\⊗\stackrel{\→}{Q}_g^b-\stackrel{\→}{s}_b---\left(6\right)$

In order to make the data sensor node attitude quaternion of estimation

Accurate as far as possible, must make function

Minimum as far as possible, namely to obtain:

In the formula, Represent four-dimensional real number field.

The optimum estimation method has a variety of, and this paper utilizes gradient descent method as the minimum estimation method.

Formula (8) has been described n interative computation and has been estimated the attitude quaternion value

Wherein μ is stride.

$\stackrel{\→}{Q}_{k+1}^g{}_b=\stackrel{\→}{Q}_k^g{}_b-\mathrm{\μ}\frac{\▿f(\stackrel{\→}{Q}_k^g{}_b,\stackrel{\→}{d}_g,\stackrel{\→}{s}_b)}{\left|\right|\▿f(\stackrel{\→}{Q}_k^g{}_b,\stackrel{\→}{d}_g,\stackrel{\→}{s}_b)\left|\right|},k=\mathrm{0,1,2}...n---\left(8\right)$

Formula (9) has represented that function f exists

Gradient on this aspect, wherein J is f pair

Jacobian.

$\▿f(\stackrel{\→}{Q}_k^g{}_b,\stackrel{\→}{d}_g,\stackrel{\→}{s}_b)=J^T(\stackrel{\→}{Q}_k^g{}_b,\stackrel{\→}{d}_g)f(\stackrel{\→}{Q}_k^g{}_b,\stackrel{\→}{d}_g,\stackrel{\→}{s}_b)---\left(9\right)$

Wherein detailed derivation does not repeat them here.Traditional gradient descent method calculates after need to doing n iteration to an estimated value, but this has increased many calculating when the calculated data sensing node attitude to making, and along with the mutual object of participation human body motion capture system increases, in the situation that the data sensor nodes increases pro rata, computer may be processed and not come over.When if this inertia motion capture system is operated on 100Hz or the higher refreshing frequency, because the feature of human motion is each

The size variation amount of value is not very large, this just allow this attitude algorithm for estimating in each renewal frequency an iteration once, as long as satisfy stride parameter μ greater than

Rate of change get final product.Then each vector observation calculates

Can be calculated by formula (10).

$\stackrel{\→}{Q}_{\▿,t}^g{}_b=\stackrel{\→}{Q}_{\mathrm{est},t-1}^g{}_b-{\mathrm{\μ}}_t\frac{\▿f}{\left|\right|\▿f\left|\right|}---\left(10\right)$

$\▿f=J_{G,B}^T(\stackrel{\→}{Q}_k^g{}_b,\stackrel{\→}{G}_g,\stackrel{\→}{B}_g)f_{G,B}(\stackrel{\→}{Q}_k^g{}_b,\stackrel{\→}{G}_g,\stackrel{\→}{a}_b,\stackrel{\→}{B}_g,\stackrel{\→}{m}_b)---\left(11\right)$

Wherein, Be upper one attitude quaternion that estimates of vector observation constantly,

The data communication device that is gathered by current time t two sensors is crossed gradient calculation and is obtained.Subscript

The computational methods of expression attitude quaternion are the vector observation.

Rational μ _tShould so that

Convergency factor greater than

Actual rate of change, but can not be too large and cause numerical value not restrained.So μ _tCan calculate according to (12).

${\mathrm{\μ}}_t=\mathrm{\α}\left|\right|\stackrel{\·}{Q}_{w,t}^g{}_b\left|\right|\mathrm{\Δt}---\left(12\right)$

Wherein, Δ t is the sampling period of sensing data;

For this cycle gyro sensor records

The value rate of change; α be one with acceleration transducer, parameter value that the magnetic flux density sensor noise is relevant.

(3), compensation blending algorithm

Final attitude is rotated estimated value

By With

Merging forms:

$\stackrel{\→}{Q}_{\mathrm{est},t}^g{}_b={\mathrm{\σ}}_t\stackrel{\→}{Q}_{\▿,t}^g{}_b+(1-{\mathrm{\σ}}_t)\stackrel{\→}{Q}_{w,t}^g{}_b---\left(13\right)$

Wherein,

Calculated by formula (4), Calculated by formula (10).σ _tWeighting parameters, optimum σ _tValue should so that

The weighting divergence equal

The weighting amount of convergence, make the attitude rotational value that calculates at last be tending towards optimum stable.In order to satisfy this condition, following equation is arranged namely:

$(1-{\mathrm{\σ}}_t)\mathrm{\β}={\mathrm{\σ}}_t\frac{{\mathrm{\μ}}_t}{\mathrm{\Δt}}---\left(14\right)$

Wherein,

Be

Convergency factor; β is gyrostatic measurement drift error (comprising high frequency period noise, quantizing noise, white Gaussian noise, coloured noise and constant value drift), namely shows as

Diverging rate, according to formula (2), can get:

$\mathrm{\β}=\left|\right|\frac{1}{2}Q\⊗\left[\begin{array}{cccc}0& w_{\mathrm{\β}}& w_{\mathrm{\β}}& w_{\mathrm{\β}}\end{array}\right]_g^b\left|\right|=\sqrt{\frac{3}{4}}{w}_{\mathrm{\β}}---\left(15\right)$

In the formula (15), Be unit quaternary element, w _βThe partially zero or temperature that is gyroscope survey is floated error amount.

Owing to will make

The weighting amount of convergence more than or equal to

The weighting divergence, then for formula (12), α does not have the upper limit, thus the hypothesis α very large, then for formula (14), can be reduced to:

${\mathrm{\σ}}_t\≈\frac{\mathrm{\β\Δt}}{{\mathrm{\μ}}_t}---\left(16\right)$

For formula (10), can be reduced to:

$\stackrel{\→}{Q}_{\▿,t}^g{}_b\≈-{\mathrm{\μ}}_t\frac{\▿f}{\left|\right|\▿f\left|\right|}---\left(17\right)$

Wushu (4), formula (15) and formula (17) substitution formula (13) can get:

$\stackrel{\→}{Q}_{\mathrm{est},t}^g{}_b=\frac{\mathrm{\β\Δt}}{{\mathrm{\μ}}_t}(-{\mathrm{\μ}}_t\frac{\▿f}{\left|\right|\▿f\left|\right|})+(1-\frac{\mathrm{\β\Δt}}{{\mathrm{\μ}}_t})(\stackrel{\→}{Q}_{\mathrm{est},t-1}^g{}_b+\stackrel{\·}{Q}_{w,t}^g{}_b\mathrm{\Δt})$

$=-\mathrm{\β\Δt}\frac{\▿f}{\left|\right|\▿f\left|\right|}+(1-\frac{\mathrm{\β\Δt}}{{\mathrm{\μ}}_t})(\stackrel{\→}{Q}_{\mathrm{est},t-1}^g{}_b+\stackrel{\·}{Q}_{w,t}^g{}_b\mathrm{\Δt})---\left(18\right)$

Because α is very large for hypothesis, then μ _tTo be very large also, then Can be reduced to:

$\stackrel{\→}{Q}_{\mathrm{est},t}^g{}_b\≈-\mathrm{\β\Δt}\frac{\▿f}{\left|\right|\▿f\left|\right|}+\stackrel{\→}{Q}_{\mathrm{est},t-1}^g{}_b+\stackrel{\·}{Q}_{w,t}^g{}_b\mathrm{\Δt}$

$=\stackrel{\→}{Q}_{\mathrm{est},t-1}^g{}_b+(\stackrel{\→}{Q}_{w,t}^g{}_b-\mathrm{\β}\frac{\▿f}{\left|\right|\▿f\left|\right|})---\left(19\right)$

So far, according to formula (19), in the cycle, the estimated value of two kinds of computation method for attitude is synthesized the attitude quaternion of cumulative upper one-period in each Data Update

Can obtain the attitude quaternion value of comprehensive two estimated values

According to the thinking of above compensation blending algorithm, based on the compensating filter structure algorithm FB(flow block) of gradient descent algorithm as shown in figure 14.By calculating the attitude quaternion value, the sensing data of original 18Byte of per cycle can be reduced to 8Byte, can significantly shorten data transmission period, increase entrained data sensor number of nodes.

6, the data of data sensor node after according to collection period encapsulation compression, and enter low-power consumption resting state waiting timer it is waken up.

The data sensor node is according to frequency acquisition multi collect sensing data, and after calculating the attitude quaternary element value of each collection, will utilize the pipeline data polymerization to come the packing Frame.Data aggregate is as the important model of wireless routing and be suggested, its main principle is exactly comprehensively from different nodes or the different data constantly of same node, by merging homogeneous data and removing redundancy, reduce volume of transmitted data, thereby save bandwidth and energy consumption.This algorithm has been realized from the conventional method centered by the address (seek shortest path by) to data-centered method the transformation of (be fixed route basically between the destination node, in the net during transmission down redundant data) between two addressable destination nodes.

The data aggregation method based on pipeline that proposes for the present invention as shown in figure 15.Detailed process is: the data sensor node periodically gathers the compressed sensing data according to the Refresh Data cycle of relay station notice, the attitude quaternion value that per cycle calculates is not sent to relay station at once, but be buffered in the inner buffer pipeline, until the data passback of appointing with relay station constantly just transfers to relay station with data.By the pipeline data polymerization, can effectively reduce the synchronization times of data sensor node and relay station, reduce synchronization frame taking wireless bandwidth.The packaged rear data sensor node of Frame will be entered the low-power consumption resting state, and waiting timer wakes it up and carries out transfer of data.

7, the data sensor node is according to the collection period timing wake-up, and sends data to the specific data via node.

After networking was finished, relay station just began the polling data sensing node, and the data sensor node calculates for the first time moment of transfer of data according to turnover rate and time slot position, then sets timer, gathered packaged data and just entered low power consumpting state afterwards.Timer can wake the data sensor node up in the moment that needs transfer of data, and the data sensor node then is transferred to packaged Frame the relay station of appointment at this transmission time slot.After transfer of data was finished, the data sensor node will be according to turnover rate, calculated next time the moment of transfer of data (during such as 30Hz, after the 33ms), and the setting timer, again gather the compression encapsulation of data, then enter the low power consumpting state waiting timer and again it is waken up.

8, convergence node receive data regeneration node data, and be sent to computer by USB.

6 data via nodes that convergence node periodic polls is managed, and by USB the new data in one cycle is sent to computer, finish the collection of an exercise data.

9, DTD, repeating step 5-8 realizes the periodicity collection of motion-sensing data.

The method for synchronizing time that this collector adopts is relay station take the network address of receiving the convergence node and distributing as time origin, after this adopt timer and dynamic calibration technology definite next time with moment of convergence node communication.After this data sensor node adopts definite moment of next time communicating by letter with relay station of timer and dynamic calibration technology take the network address of receiving relay station and distributing as time origin.

Time error between the such Time Synchronization Mechanism, itself and relay station can not surpass a update cycle, and the accumulative total effect can not occur yet.Precision of timer between data sensor node and the relay station is poor less, and the data sensor node is in time of low power consumpting state just can be longer, also just more saves energy, otherwise the data sensor node need to wake up in advance, prevents from missing polling time slot.

Time Synchronization Mechanism between relay station and the convergence node similarly, relay station is to finish the moment of a communication process (networking or transfer of data) as time origin with the convergence node, calculate the moment of next time communicating by letter according to time slot position with turnover rate, set timer.Finish the communication process with the data sensor node, waiting timer is overtime, again with the convergence node communication.

The time slot allocation of whole system as shown in figure 16.The synchronous promoter of system is the convergence node, in each Frame the cycle, it is by timesharing (time division multiple access, TDMA) form and each relay station communicate, each relay station is used different radio-frequency carrier channels, can work simultaneously, relay station is also passed through the format management sensor node separately of TDMA.Can set up the time synchronized of the radio sensing network of whole sensing data acquisition system by this time slot allocation relation.

The topology of this radio sensing network is relatively-stationary behind system stable operation: each data sensor node and relay station single-hop communication, each relay station and the direct communication of convergence node.The concrete software flow pattern of data sensor node, relay station and convergence node is respectively such as Figure 17, and Figure 18 is shown in Figure 19.

Above-described embodiment is used for the present invention that explains, rather than limits the invention, and in the protection range that spirit of the present invention and right are asked, any modification and change to the present invention makes all fall into protection scope of the present invention.

Claims (10)

1. wireless many people motion data collection element, it is characterized in that, it is mainly by dozens of data sensor node, six data via nodes and a data aggregation node form, three kinds of nodes form radio sensing network based on sub-clustering in order to accurately, gather at high speed every human motion sensing data; Wherein, dozens of data sensor node is attached to each motion key position of human body, after the data compression that periodically gathers, be sent to the relay station that connects by wireless network, convergence node and six data via nodes are by SPI interface wired connection, tdm communication, sensing data gathers to computer by USB the most at last.

2. many people motion data collection element data sensor node according to claim 1, it is characterized in that: described data sensor node mainly comprises: MCU+ wireless radio frequency modules, acceleration sensor module, magnetic induction sensor module, MEMS numeral output gyroscope modules and power supply conditioning and monitoring modular etc.; The MCU+ wireless radio frequency modules is connected with plate and carries antenna, and MCU passes through I ²The C bus links to each other with MEMS numeral output gyroscope modules, magnetic induction sensor module, links to each other with acceleration sensor module by the SPI interface; The data sensor node produces the supply power voltage that each hardware module needs by the power supply of 3.3V/250mAH chargeable lithium cell through the power supply conditioning module, the electric weight of power supply monitoring module monitors battery, and it is by common IO mouth Simulation with I ²The C bus protocol is connected with MCU.

3. many people motion data collection element relay station according to claim 1 and convergence node is characterized in that: described relay station and convergence node main task are management sensing network, forward command and forwarding datas; Large for the collector data transfer rate, real-time characteristics, relay station and convergence node adopt SPI interface wired connection, and relay station and convergence node are integrated on the circuit board, by same power module power supply; The convergence node is as the interface of whole data acquisition system and computer, its task is by 6 data via nodes of SPI interface management, and by USB interface forwarding computer command and transmission human body movement data, it mainly comprises two modules: arm processor module and usb interface module; Collector has 6 data via nodes, and each relay station all can be managed and is no more than 8 wireless data sensing nodes, and it is responsible for processing and the forwarding of motion-sensing data, and it mainly comprises two modules: dsp processor module and radio receiving transmitting module; Dsp processor is connected with radio receiving transmitting module by the SPI mouth, and communicates by letter with arm processor by another SPI interface.

4. the many people exercise data acquisition method based on the clustering wireless sensor network network is characterized in that, may further comprise the steps:

(1) convergence node and relay station initialization;

(2) radio sensing network is set up in convergence node and relay station networking, waits for that the data sensor node networks;

(3) data sensor node initializing;

(4) data sensor node monitor channel, application networks;

(5) data sensor node periodicity pick-up transducers data, and process this secondary data of compression;

(6) data of data sensor node after according to collection period encapsulation compression, and enter low-power consumption resting state waiting timer it is waken up;

(7) the data sensor node is according to the collection period timing wake-up, and sends data to the specific data via node;

(8) convergence node receive data regeneration node data, and be sent to computer by USB;

(9) DTDs, repeating step 5-8 realizes the periodicity collection of motion-sensing data.

5. many people exercise data acquisition method according to claim 4, it is characterized in that, described step (1) is specially: convergence node and relay station initialization comprise the initialization of arm processor, the initialization of dsp processor, radio-frequency module initialization; The initialization of arm processor and the initialization of dsp processor include bus frequency and set, the configuration of GPIO mouth, interrupt configuration, SPI interface configuration etc.; The radio-frequency module initialization comprises channel fix, clock frequency setting etc.; Described step (3) is specially: the data sensor node initializing comprises the initialization of MCU+ wireless radio frequency modules, acceleration sensor module initialization, magnetic induction sensor module initialization, the initialization of MEMS numeral output gyroscope modules; The initialization of MCU+ wireless radio frequency modules comprises the configuration of GPIO mouth, interrupt configuration, channel fix, clock frequency setting etc.; Three kinds of sensor die initialization blocks comprise setting state, data initialization etc.

6. many people exercise data acquisition method according to claim 4, it is characterized in that, described step (2) is specially: the convergence node is according to computer command setting network refreshing frequency, and periodically send the networking beacon and communicate by letter with relay station by the SPI interface, carry out the ground floor networking of radio sensing network, comprise following substep:

1) convergence node broadcasts networking commands;

2) relay station log on address;

3) information such as convergence node distribution network address, time slot position and configurating channel;

4) the convergence node is confirmed, sets up radio sensing network;

5) relay station periodic broadcast networking commands waits for that the data sensor node networks.

7. many people exercise data acquisition method according to claim 4 is characterized in that, described step (4) is specially: data sensor node and relay station form the second layer of this collector wireless sensing collection network; Relay station periodic broadcast networking commands, the data sensor node is selected suitable relay station, to its log on address, mainly comprises following substep:

1) relay station broadcast networking order;

2) the data sensor node is searched for legal relay station according to back off algorithm, and to its log on address;

3) information such as relay station distribution network address, time slot position and turnover rate;

4) the data sensor node is confirmed the relay station response message, and successfully networking then enters next step, and is failed then carry out back off algorithm, iteron step 2)-4);

5) enter the data acquisition transmit stage.

8. many people exercise data acquisition method according to claim 4 is characterized in that, described step (5) is specially: the data sensor node passes through I according to the frequency acquisition of relay station notice ²C bus and SPI interface gather three kinds of sensing datas of 9 axle 18Byte, and utilize local attitude quaternion filtering method of estimation to calculate the attitude quaternion of this image data, reach the data compression purpose by transmitting 8Byte four element datas.

9. many people exercise data acquisition method according to claim 4, it is characterized in that, described step (6) is specially: after sensing data is calculated in the collection of data sensor node, can be according to the Data Update cycle of relay station notice, the attitude quaternion value that a plurality of calculating are good is buffered in the inner buffer pipeline to utilize the pipeline data polymerization, the packaged rear data sensor node of Frame will enter the low-power consumption resting state, and waiting timer wakes it up and carries out transfer of data again.

10. many people exercise data acquisition method according to claim 2, it is characterized in that, described step (7) is specially: after networking is finished, relay station just begins the polling data sensing node, and the data sensor node can calculate for the first time moment of transfer of data according to turnover rate and time slot position, set timer, this timer can wake the data sensor node up in the moment that needs transfer of data, and the data sensor node is then with the relay station of packaged data frame transfer to appointment; Described step (8) is specially: 6 data via nodes that convergence node periodic polls is managed, and by USB the new data in one cycle is sent to computer, finish the collection of an exercise data.

Images