CN105180959A - Anti-interference step counting method for wrist type step counting devices - Google Patents

Abstract

The present invention discloses an anti-interference step counting method for wrist type step counting devices. According to the present invention, the analysis on the relationship between various peak values of the frequency domain is used to replace the method for carrying out wave filtering and smoothing treatment on the collected acceleration information and then calculating in the prior art so as to improve the step counting efficiency; and according to the set threshold parameter, the method using the frequency domain calculation as the main and using the time domain calculation as the reference is used to perform the step counting, such that the compensation can be performed through the time domain calculation result under the inaccurate frequency domain calculation result so as to ensure the accuracy of the algorithm as far as possible.

Description

A kind of anti-interference step-recording method being applicable to wrist-passometer

Technical field

The present invention relates to consumer applications electronic technology field, be specifically related to a kind of anti-interference step-recording method being applicable to wrist-passometer.

Background technology

Passometer is mainly used in monitoring the daily amount of exercise of people, and user can be helped to a certain extent to build up health.

At present, the acceleration information produced when passometer adopts 3-axis acceleration sensor to gather human body walking usually, extracts correlated characteristic wherein, as frequency, peak value, slope etc.This kind of step-recording method has following shortcoming usually:

1, different user's movement acceleration information differs greatly, in order to choose suitable threshold value, usually can carry out filtering, smoothing processing to the acceleration information collected in prior art, do like this and can consume more hardware resource, bring certain computation delay problem simultaneously;

2, existing frequency domain step-recording method normally by the acceleration information that collects after Fourier transform, the frequency directly chosen corresponding to spectral magnitude maximum point walks result as meter, but wrist-passometer introduces larger noise due to unordered the rocking of human body wrist in motion process usually, the spectral magnitude of noise often exceeds actual motion spectral magnitude, cause information extraction mistake, further impact meter step result.

Summary of the invention

In view of this, the invention provides a kind of anti-interference step-recording method being applicable to wrist-passometer, the accuracy that motion frequency extracts can be improved.

Be applicable to an anti-interference step-recording method for wrist-passometer, its concrete steps are as follows:

Step one, sampling interval according to setting, three direction of principal axis acceleration information during human body walking press three-dimensional cartesian coordinate system decompose, acquisition 3-axis acceleration information a _x, a _yand a _z, and resultant acceleration a _s; Gather N time altogether, draw about horizontal ordinate be sampled point, ordinate is the resultant acceleration oscillogram of resultant acceleration value;

Step 2, utilize resultant acceleration oscillogram, determine effective peak point, Time Domain Amplitude corresponding for each effective peak point and sampled point are formed set A; Obtain Time Domain Amplitude threshold value T _ath, and according to Time Domain Amplitude threshold value T _athdetermine that set A Time Domain Amplitude is greater than Time Domain Amplitude threshold value T _athsampled point, formed set B; Obtain sampling interval threshold value T _tth, and according to sampling interval threshold value T _tthdetermine satisfactory sampled point in set B, form set C; The last step number result Q obtained according to Q=2 × (l-1) under time domain; Wherein, l is the number of sampled point in set C;

Step 3, according in step one obtain resultant acceleration a _s, by taking absolute value after Fast Fourier Transform (FFT), obtain resultant acceleration spectrogram;

Step 4, resultant acceleration spectrogram carried out to first time Frequency point screening:

401st step: the spectral magnitude being approximately the frequency point value place of zero carries out return-to-zero;

402nd step: in resultant acceleration spectrogram, determines effective peak point, and frequency domain amplitude corresponding for each effective peak point and Frequency point are formed set D;

Step 5, determine to gather the frequency domain amplitude threshold value P that D frequency domain amplitude is more than or equal to setting _athfrequency point, formed set E, complete programmed screening;

Step 6, by set E in all Frequency points with approximate frequency multiplication relation screen, formed set F, complete third time screening;

Step 7, carry out the initial setting of weights to all Frequency points in set F, frequency values is pressed order arrangement from small to large, and be set to the highest by the weights corresponding to Frequency point minimum for frequency point value, be set as q, wherein, q is more than or equal to Frequency point number; And by q=q-1, according to the order of sequence weight setting is carried out to other Frequency points;

Step 8, according in step one obtain 3-axis acceleration information, draw the acceleration spectrogram of three axles respectively, according to the method for step 4, the spectral magnitude being approximately the frequency point value place of zero carries out return-to-zero; And determine effective peak point, obtain Frequency point corresponding to each peak point and corresponding frequency domain amplitude, form 3 corresponding set; Afterwards, according to the method for step 5, for 3 set, determine its frequency domain amplitude threshold value respectively, and the Frequency point being greater than respective frequency domain amplitude threshold value in each set is filtered out, form 3 corresponding new set; For each described new set, according to the method for step 6, obtain the Frequency point with approximate frequency multiplication relation respectively, finally form set G respectively ₁, G ₂and G ₃;

Step 9, for set G ₁, G ₂and G ₃in any one set, for any one Frequency point O, in set E, search the Frequency point whether having and differ 2 with the position of this Frequency point O, if had, then by set E accordingly the weights of Frequency point add 1;

Step 10, the Frequency point of maximum weight in set E to be screened, as present frequency point, and carry out the corresponding frequency values f of conversion acquisition _p;

Step 11, according to the frequency f obtained in step 10 _p, obtain the step number Q ' under frequency-domain result;

Whether the step number result Q obtained in step 12, determining step two becomes multiple proportion with the step number Q ' obtained in step 11; If one-tenth multiple proportion, then perform step 13; Otherwise, the step number obtained in step 11 is exported as final meter step result;

Step 13, according to formula Q "=Q × H ₁+ Q' × H ₂, obtain final fitting result Q ", and export as final meter step result.Wherein, H ₁for the time domain fitting coefficient obtained after carrying out matching for true step number and true time-domain step number measurement result in advance; H ₂for the Frequency Fitting coefficient obtained after true step number and true frequency domain step number measurement result carry out matching in advance.

Especially, the method obtaining Time Domain Amplitude threshold value in step 2 is: according to resultant acceleration oscillogram, by Time Domain Amplitude maximum of T _amaxwith Time Domain Amplitude minimum value T _aminintermediate value as Time Domain Amplitude threshold value T _ath.

Especially, time-domain sampling interval threshold T is utilized in step 2 _tth, determine that the method for satisfactory sampled point in set B is:

Using the sampled point of first in set B as reference point, judge whether the sampling interval be adjacent between sampled point is greater than threshold value T _tth:

If be more than or equal to, current base point is saved to set C in, and then using its neighbouring sample point as new reference point; Once compare on carrying out in set B, in traversal set B till all sampled points;

If be less than, judge the amplitude size of current base point and neighbouring sample point, using sampled point large for amplitude as new reference point, judge whether the sampling interval be adjacent between sampled point is greater than threshold value T _tth, in traversal set B till all sampled points.

Especially, step 5 frequency domain amplitude thresholds P _athpreparation method be: will set D frequency domain amplitude minimum value and the intermediate value of maximal value as frequency domain amplitude threshold value P _ath.

Especially, when base times frequency and the power error doubly between frequency are less than or equal to 2, then think that there is approximate frequency multiplication relation.

Especially, sampling interval threshold value T in step 2 _tththe minimum amplitude limit numerical value in the different arms swing cycles corresponding to the people of all ages and classes or walking habits, namely using 0.2 second to 0.4 second as the minimum clipping range of universality, and according to

$T_{tth}=\left\{\begin{array}{cc}0.4,& \frac{g}{T_{ath\_norm}}>0.4\\ \frac{0.2}{T_{ath\_norm}},& 0.2\&le;\frac{g}{T_{ath\_norm}}\&le;0.4\\ 0.2,& \frac{g}{T_{ath\_norm}}<0.2\end{array}\right.,$ Obtain sampling interval threshold value T _tth; G is fitting constant; T _{ath_norm}for amplitude thresholds normalized in set B.

Beneficial effect:

Can produce harmonic wave due to during human body walking swing arm, therefore the present invention is by the relation analyzed between each peak value of frequency domain, instead of the method calculated again after carrying out filtering, smoothing processing to the acceleration information collected in prior art, improves meter step efficiency.In addition, because the present invention is according to the threshold parameter of setting, based on frequency-domain calculations, time-domain calculation method as a reference carries out meter step, under the result that frequency-domain calculations result is inaccurate, can be compensated by time-domain calculation result, ensure algorithm accuracy as much as possible.

Frequency-domain calculations utilizes the harmonic characteristic of arms swing to carry out frequency computation part, once frequency domain meter step result and time domain meter to walk result different, just obtaining net result by the mode of matching, also can ensure the accuracy that motion frequency extracts as much as possible when disturbing larger like this.

Accompanying drawing explanation

Fig. 1 is the system flowchart of anti-interference step-recording method of the present invention.

Fig. 2 is resultant acceleration oscillogram and corresponding Time Domain Amplitude threshold value and time-domain sampling interval threshold.

Fig. 3 is resultant acceleration spectrogram and corresponding frequency domain amplitude thresholds and times frequency

Embodiment

To develop simultaneously embodiment below in conjunction with accompanying drawing, describe the present invention.

The invention provides a kind of anti-interference step-recording method being applicable to wrist-passometer: its main thought is: can harmonic wave be produced according to during human body walking swing arm, by the relation analyzed between frequency domain each peak value, and then obtain frequency domain note step result.

In addition, because the present invention is according to the threshold parameter of setting, based on frequency-domain calculations, time-domain calculation method as a reference carries out meter step, by the comparison to the step number obtained under comparing frequency domain situation and time domain situation, obtains meter more accurately and walks result.In addition, frequency-domain and time-domain is combined, under the result that frequency-domain calculations result is inaccurate, can be compensated by time-domain calculation result, ensure algorithm accuracy as much as possible.

Meanwhile, frequency-domain calculations no longer adopts the method directly choosing frequency spectrum maximal value respective frequencies to carry out meter step, but utilizes the harmonic characteristic of arms swing to carry out frequency computation part, also can ensure the accuracy that motion frequency extracts as much as possible like this when disturbing larger.

As shown in Figure 1:

Step one, according to the sampling interval of setting and sample frequency f, acceleration information during human body walking press three direction of principal axis of three-dimensional cartesian coordinate system and X-direction, Y direction and Z-direction are decomposed, acquisition 3-axis acceleration information a _x, a _yand a _z.And according to

$a_s=\sqrt{{a_x}^2+{a_y}^2+{a_z}^2}---\left(1\right),$

Obtain the resultant acceleration a of current sampling point _s.Altogether sample N time, and correspondence by the resultant acceleration a of each sampled point _sdraw resultant acceleration oscillogram, wherein, horizontal ordinate is sampled point, and ordinate is resultant acceleration value and Time Domain Amplitude T.As shown in Figure 2

Step 2,

1st step: utilize resultant acceleration oscillogram, carry out first time sampled point screening:

According to resultant acceleration oscillogram, utilize method of difference to determine effective peak point, Time Domain Amplitude corresponding for each effective peak point and sampled point are formed set A.

2nd step: for the sampled point with Time Domain Amplitude information in set A, carry out the screening of second time sampled point:

By the feature extraction of involutory Acceleration pulse figure, determine the Time Domain Amplitude threshold value T of each sampled point _ath: the Time Domain Amplitude T namely extracting each sampled point in set A; Therefrom filter out amplitude maximum T _amaxwith amplitude minimum value T _amin, and according to

$T_{ath}=\frac{1}{2}(T_{amax}+T_{a\min })---\left(2\right),$

Obtain Time Domain Amplitude threshold value T _ath.And in set A, find out Time Domain Amplitude be greater than Time Domain Amplitude threshold value T _athsampled point, formed set B.

3rd step, for each sampled point in set B, carry out third time sampled point screening:

231 steps: for each sampled point in set B, by the feature extraction of resultant acceleration oscillogram, determine time-domain sampling interval threshold T _tth.That is: according to human body characteristic analysis when walking, general, the minimum amplitude limit numerical value in the different arms swing cycles corresponding to the people of all ages and classes or walking habits, such as, older elder or walking more slowly people when walking, its swing arm cycle is not less than 0.4 second; And young man or walk that the swing arm cycle is not less than 0.2 second to people faster when walking; For this reason, the present invention is using 0.2 second to 0.4 second as the minimum clipping range of universality.In addition, in order to improve the accuracy of result, as much as possible the sampled point in set A is analyzed, the present invention is according to resultant acceleration and the relation in arms swing cycle: when resultant acceleration value is larger, then illustrate that user's speed of travel is faster, now, the cycle of its arms swing is then faster, and Time Domain Amplitude is also larger; Therefore sampling interval threshold value should reduce accordingly.The present invention proposes piecewise function formula (3)

$T_{tth}=\{\begin{array}{cc}0.4,& \frac{g}{T_{ath\_norm}}>0.4\\ \frac{0.2}{T_{ath\_norm}},& 0.2\&le;\frac{g}{T_{ath\_norm}}\&le;0.4\\ 0.2,& \frac{g}{T_{ath\_norm}}<0.2\end{array}---\left(3\right),$

Wherein, g is fitting constant; Usually, g gets 0.2.Tath_norm is normalized amplitude thresholds in set B, and:

$T_{ath\_norm}=\frac{T_{ath}}{T_{max}}---\left(4\right),$

Wherein, T _maxfor Time Domain Amplitude maximal value in set B.

232 steps: search the sampled point meeting sampling interval threshold requirement.Using the sampled point of first in set B as reference point, judge whether the sampling interval be adjacent between sampled point is greater than threshold value T _tth:

If be more than or equal to, current base point is saved to set C in, and then using its neighbouring sample point as new reference point; Once compare on carrying out in set B, in traversal set B till all sampled points;

If be less than, judge the amplitude size of current base point and neighbouring sample point, using sampled point large for amplitude as new reference point, judge whether the sampling interval be adjacent between sampled point is greater than threshold value T _tth, in traversal set B till all sampled points.

233 steps: all reference points found are formed new set C; Number l in statistics set C.Now, in set C, two adjacent sampled points then think the one-period of arms swing, therefore obtain step number result Q in the time domain:

Q＝2×(l-1)(5)，

Wherein, (l-1) represents periodicity.Because time-domain signal is subject to external interference relatively; when particularly wrist disturbance is comparatively strong, often there will be the problem of calculation deviation, therefore in step-recording method of the present invention; need the result of time domain acquisition and frequency domain to contrast, and the result that time domain obtains is used for supplementary frequency-domain step number result.

Step 3, rendering accelerating degree spectrogram:

According to the resultant acceleration a obtained in step one _s, by taking absolute value after Fast Fourier Transform (FFT), obtaining resultant acceleration frequency spectrum, and drawing resultant acceleration spectrogram; Wherein, horizontal ordinate represents Frequency point, and ordinate represents this Frequency point place spectral magnitude.Frequency point value increases from left to right successively, and Frequency point passes through formula

$f_P=\frac{J}{K}\&times;f---\left(6\right),$

, the frequency of user's arms swing when walking can be obtained, namely represent the number of times of arms swing in the unit interval.Wherein, J represents frequency point value, and K represents fast Fourier sampling number, and f is sampling interval.Therefore in practical situations both, the frequency of arms swing p.s. is very slow, be generally no more than 1Hz to 5Hz, and frequency point value is directly proportional to frequency, therefore when analyzing, only need the spectrogram paying close attention to low frequency part.

Step 4, resultant acceleration spectrogram carried out to first time Frequency point screening:

401st step: due in the process that changes in human body swing arm attitude, Acceleration pulse can produce certain shake, i.e. baseline wander, and the frequency of this shake is lower, and frequency point value is approximately zero.And after carrying out Fast Fourier Transform (FFT), larger amplitude is had near the Frequency point that frequency point value is approximately zero, this can produce larger impact for follow-up spectral magnitude threshold process, and for this reason, the present invention has first carried out return-to-zero to the spectral magnitude of this part.Wherein, the Frequency point that frequency point value is here approximately zero refers to the Frequency point of frequency corresponding to 0Hz to 0.2Hz

402nd step: in resultant acceleration spectrogram, utilizes method of difference to determine effective peak point, and frequency domain amplitude corresponding for each effective peak point and Frequency point are formed set D;

Step 5, for set D in each Frequency point with frequency domain amplitude carry out second time Frequency point screen:

For each Frequency point in set D, filter out amplitude maximum P _amaxwith amplitude minimum value P _amin, and according to formula (6)

$P_{ath}=\frac{1}{2}(P_{amax}+P_{a\min })---\left(7\right),$

Obtain frequency domain amplitude threshold value P _ath.And find out and be greater than frequency domain amplitude threshold value P in set D _athfrequency point, formed set E, complete programmed screening.

Step 6, for set E in each Frequency point carry out third time Frequency point screen:

Filter out the Frequency point in set E with approximate frequency multiplication relation; As shown in Figure 3, two peak points marked in figure are all on threshold value, and corresponding frequency point value is approximated to 2 times of relations, and at this moment, first point is called fundamental frequency, second point becomes a times frequency.Due in real process, the noisy existence of meeting in real resultant acceleration, so a times frequency here refers to approximate times frequency.According to error requirements, its base times frequency and the error number doubly between frequency are less than or equal to 2.All Frequency points with frequency multiplication relation in set E are screened, forms set F, complete third time screening.

Step 7, due in actual human body walking process, the frequency of human body swing arm is slower, therefore to set F in all Frequency points carry out the initial setting of weights time, frequency values is pressed order arrangement from small to large, and the weights corresponding to Frequency point minimum for frequency point value are set to the highest, be set as q, wherein, q is more than or equal to Frequency point number; And by q=q-1, according to the order of sequence weight setting is carried out to other Frequency points.To make to obtain more actual frequency value comparatively accurately when the later stage screens weights.Such as, the Frequency point arranged from small to large for 1,2,3,4,5,6,7,8}, and definition frequency point value be 1 weights be 10, then { weights corresponding to 1,2,3,4,5,6,7,8} are { 10,9,8,7,6,5,4,3, } to Frequency point;

Step 8, according in step one obtain 3-axis acceleration information, draw the acceleration spectrogram of X-axis, Y-axis and Z axis respectively, according to the method for step 4, the spectral magnitude being approximately the frequency point value place of zero carries out return-to-zero; And determine effective peak point, obtain Frequency point corresponding to each peak point and corresponding frequency domain amplitude, form 3 corresponding set; Afterwards, according to the method for step 5, for 3 set, determine its frequency domain amplitude threshold value respectively, and the Frequency point being greater than respective frequency domain amplitude threshold value in each set is filtered out, form 3 corresponding new set; For each described new set, according to the method for step 6, obtain the Frequency point with approximate frequency multiplication relation respectively, finally form set G respectively ₁, G ₂and G ₃.

Step 9, for set G ₁, G ₂and G ₃in any one set, for any one Frequency point O, in set E, search the Frequency point whether having and differ 2 with the position of this Frequency point O, if had, then by set E accordingly the weights of Frequency point add 1.

Step 10, the Frequency point of maximum weight in set E to be screened, as present frequency point, and according to formula (6), Frequency point is converted to frequency values f _p.

Step 11, according to the frequency f obtained in step 10 _p, and then directly obtain swing arm periodicity; And according to formula (5), obtain the step number Q ' under frequency-domain result.

Whether the step number result Q obtained in step 12, determining step two becomes multiple proportion with the step number Q ' obtained in step 11; If one-tenth multiple proportion, then perform step 13; Otherwise, the step number result obtained in step 10 is exported as final meter step result.

A frequency multiplication or two frequency multiplication positions are appeared in order to prevent the harmonic wave produced during human body swing arm in the process of walking, and then cause the Frequency point that filters out at frequency domain wrong, the step number Q ' that the present invention proposes obtaining in the step number result Q step 11 obtained in step 2 judges, check whether it becomes multiple proportion, if one-tenth multiple proportion, then illustrate that Frequency point is selected in humorous wave point position, then now need to carry out matching to frequency domain step number result and time domain step number result.Namely step 13 is performed.

Step 13, basis

Q”＝Q×H ₁+Q'×H ₂(8)，

Obtain final fitting result Q ", and export as final meter step result.Wherein, H ₁for the time domain fitting coefficient obtained after carrying out matching for true step number and true time-domain step number measurement result in advance; H ₂for the Frequency Fitting coefficient obtained after true step number and true frequency domain step number measurement result carry out matching in advance.

In sum, these are only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (6)

1. be applicable to an anti-interference step-recording method for wrist-passometer, it is characterized in that, its concrete steps are as follows:

Step one, sampling interval according to setting, three direction of principal axis acceleration information during human body walking press three-dimensional cartesian coordinate system decompose, acquisition 3-axis acceleration information a _x, a _yand a _z, and resultant acceleration a _s; Gather N time altogether, draw about horizontal ordinate be sampled point, ordinate is the resultant acceleration oscillogram of resultant acceleration value;

Step 2, utilize resultant acceleration oscillogram, determine effective peak point, Time Domain Amplitude corresponding for each effective peak point and sampled point are formed set A; Obtain Time Domain Amplitude threshold value T _ath, and according to Time Domain Amplitude threshold value T _athdetermine that set A Time Domain Amplitude is greater than Time Domain Amplitude threshold value T _athsampled point, formed set B; Obtain sampling interval threshold value T _tth, and according to sampling interval threshold value T _tthdetermine satisfactory sampled point in set B, form set C; The last step number result Q obtained according to Q=2 × (l-1) under time domain; Wherein, l is the number of sampled point in set C;

Step 3, according in step one obtain resultant acceleration a _s, by taking absolute value after Fast Fourier Transform (FFT), obtain resultant acceleration spectrogram;

Step 4, resultant acceleration spectrogram carried out to first time Frequency point screening:

401st step: the spectral magnitude being approximately the frequency point value place of zero carries out return-to-zero;

402nd step: in resultant acceleration spectrogram, determines effective peak point, and frequency domain amplitude corresponding for each effective peak point and Frequency point are formed set D;

Step 5, determine to gather the frequency domain amplitude threshold value P that D frequency domain amplitude is more than or equal to setting _athfrequency point, formed set E, complete programmed screening;

Step 6, by set E in all Frequency points with approximate frequency multiplication relation screen, formed set F, complete third time screening;

Step 7, carry out the initial setting of weights to all Frequency points in set F, frequency values is pressed order arrangement from small to large, and be set to the highest by the weights corresponding to Frequency point minimum for frequency point value, be set as q, wherein, q is more than or equal to Frequency point number; And by q=q-1, according to the order of sequence weight setting is carried out to other Frequency points;

Step 8, according in step one obtain 3-axis acceleration information, draw the acceleration spectrogram of three axles respectively, according to the method for step 4, the spectral magnitude being approximately the frequency point value place of zero carries out return-to-zero; And determine effective peak point, obtain Frequency point corresponding to each peak point and corresponding frequency domain amplitude, form 3 corresponding set; Afterwards, according to the method for step 5, for 3 set, determine its frequency domain amplitude threshold value respectively, and the Frequency point being greater than respective frequency domain amplitude threshold value in each set is filtered out, form 3 corresponding new set; For each described new set, according to the method for step 6, obtain the Frequency point with approximate frequency multiplication relation respectively, finally form set G respectively ₁, G ₂and G ₃;

Step 9, for set G ₁, G ₂and G ₃in any one set, for any one Frequency point O, in set E, search the Frequency point whether having and differ 2 with the position of this Frequency point O, if had, then by set E accordingly the weights of Frequency point add 1;

Step 10, the Frequency point of maximum weight in set E to be screened, as present frequency point, and carry out the corresponding frequency values f of conversion acquisition _p;

Step 11, according to the frequency f obtained in step 10 _p, obtain the step number Q ' under frequency-domain result;

Whether the step number result Q obtained in step 12, determining step two becomes multiple proportion with the step number Q ' obtained in step 11; If one-tenth multiple proportion, then perform step 13; Otherwise, the step number obtained in step 11 is exported as final meter step result;

Step 13, according to formula Q "=Q × H ₁+ Q' × H ₂, obtain final fitting result Q ", and export as final meter step result.Wherein, H ₁for the time domain fitting coefficient obtained after carrying out matching for true step number and true time-domain step number measurement result in advance; H ₂for the Frequency Fitting coefficient obtained after true step number and true frequency domain step number measurement result carry out matching in advance.

2. anti-interference step-recording method as claimed in claim 1, is characterized in that: the method obtaining Time Domain Amplitude threshold value in step 2 is: according to resultant acceleration oscillogram, by Time Domain Amplitude maximum of T _amaxwith Time Domain Amplitude minimum value T _aminintermediate value as Time Domain Amplitude threshold value T _ath.

3. anti-interference step-recording method as claimed in claim 1, is characterized in that: utilize time-domain sampling interval threshold T in step 2 _tth, determine that the method for satisfactory sampled point in set B is:

Using the sampled point of first in set B as reference point, judge whether the sampling interval be adjacent between sampled point is greater than threshold value T _tth:

If be more than or equal to, current base point is saved to set C in, and then using its neighbouring sample point as new reference point; Once compare on carrying out in set B, in traversal set B till all sampled points;

If be less than, judge the amplitude size of current base point and neighbouring sample point, using sampled point large for amplitude as new reference point, judge whether the sampling interval be adjacent between sampled point is greater than threshold value T _tth, in traversal set B till all sampled points.

4. anti-interference step-recording method as claimed in claim 1, is characterized in that: step 5 frequency domain amplitude thresholds P _athpreparation method be: will set D frequency domain amplitude minimum value and the intermediate value of maximal value as frequency domain amplitude threshold value P _ath.

5. anti-interference step-recording method as claimed in claim 1, is characterized in that: when base times frequency and the power error doubly between frequency are less than or equal to 2, then think to have approximate frequency multiplication relation.

6. anti-interference step-recording method as claimed in claim 1, is characterized in that: sampling interval threshold value T in step 2 _tththe minimum amplitude limit numerical value in the different arms swing cycles corresponding to the people of all ages and classes or walking habits, namely using 0.2 second to 0.4 second as the minimum clipping range of universality, and according to

$T_{tth}=\left\{\begin{array}{cc}0.4,& \frac{g}{T_{ath\_norm}}>0.4\\ \frac{0.2}{T_{ath\_norm}},& 0.2\&le;\frac{g}{T_{ath\_norm}}\&le;0.4\\ 0.2,& \frac{g}{T_{ath\_norm}}<0.2\end{array}\right.,$ Obtain sampling interval threshold value T _tth; G is fitting constant; T _{ath_norm}for amplitude thresholds normalized in set B.