CN106250435A - A kind of user's scene recognition method based on mobile terminal Noise map - Google Patents

Abstract

The invention discloses a kind of user's scene recognition method based on mobile terminal Noise map, intelligent movable mobile phone is used to gather sensing data, reference voice data easily by friction and the feature of vibration influence, carry out coarseness division to the scene of user's acquisition noise decibel.Use semi-artificial notation methods to obtain test data set, be calculated sensor scene judgment threshold for different sensors feature, and use mobile terminal scene Recognition algorithm that the decibel data record gathered under different scenes is carried out scene classification.User's scene recognition method based on mobile terminal Noise map that the present invention provides, calculate and speculate that user gathers a decibel data acquisition scenarios, by analyzing this data recording quality, can aid in the raising visual quality of data of Noise map and user's incentive system, thus encourage more people to participate in community or even The Surroundings in Cities monitoring task.

Description

A kind of user's scene recognition method based on mobile terminal Noise map

Technical field

The present invention relates to a kind of user's scene recognition method based on mobile terminal Noise map, belong to group in sensing network User's scene Recognition technical field of intelligence perception.

Background technology

Intelligent perception, uses the strength of mass-rent to collect, share the physical environmental data of collection more efficiently, and based on this Research and develop more towards environment monitoring and the creative service of intelligent city with application.There are three big features: 1) low cost, aspiration The citizen participated in serve as human observer, and the environmental data gathered by contribution mobile phone saves human resources and cost of equipment maintenance；2) Map rejuvenation is fast, and stochastic and dynamic monitoring in civic makes data renewal speed accelerate, and has the most ageing；3) data are common Enjoying, participating user is the features such as the contributor of noise data.When a large amount of citizen participate in community or even The Surroundings in Cities, social activity In monitoring task, and coordinate solution timely and effectively, it is possible to use the strength of mass-rent to be devoted to environmental monitoring, safety in advance Alert, real time medical such as is helped each other at the innovative application.Although intelligent perception has broad application prospects and advantage, but in actual deployment and In application, still suffer from some technological difficulties and challenge.First the hardware of different intelligent mobile phone has different, and user uses hands There is some difference for the data that machine is collected, and secondly the purpose mode participating in collecting of user has uncertainty, furthermore how to carry Encouraging more users to participate in intelligent perception application for incentive policy, the collection efficiency and the quality that improve data are also to need to solve Problem.Herein for based on mobile terminal Noise map apply in user data gather purpose, quality, measurement scene the most true Qualitative, propose to use user's scene Recognition algorithm that user contributes the scene of sensing data classify.Gathered by effective Scene classification, the quality of data identify, it will help manager carries out data prediction, data filtering, use valid data are carried out Visualization, and formulate the work such as suitable user's incentive policy.

Classifying in order to user based on intelligent perception is gathered scene, domestic and international researcher proposes multiple identification and calculates Method, pertinent literature is as follows:

1,2010, Nicolas et al. was at " Participatory noise pollution monitoring using Mobile phones " propose to use the mode of automatic scene label, use the mode of timestamp and the manual label of user to single The record measured carries out scene identity.In the method, scene coarseness is divided into 4 classes: geographical position (city, street), time Stamp (working day, weekend), weather and user behavior (static, mobile).

2,2013, Rajib et al. was at " Ear-Phone:A context-aware noise mapping using Smart phones " scene excavation algorithm is proposed, use three axis accelerometer, preposition optical sensor that the collection position of mobile phone is entered Go coarseness division: in hands, in Bao Zhong, pocket.Researcher analysis is also extracted accelerometer and preposition optical sensor three The feature of individual different scene, and use the kNN algorithm accelerometer data to collecting carry out characteristic matching and classify, by sentencing Disconnected preposition optical sensor has unobstructed thus judges whether mobile phone is in closed state.

3,2015, Zhu caused to propose to use in " participatory aware platform data acquisition key technology research and realization " Family gathers scene atmosphere four class, i.e. home scenarios, operative scenario, Outdoor Scene and automobile scene.And use user geographical position Location and kinestate, timestamp, user terminal network connection state and historical record speculate the collection scene of user, by meter Calculate Wi-Fi and connect number and historical record, thus it is speculated that user is in the probability of each scene.

Summary of the invention

Problem to be solved by this invention is: in Noise map based on intelligent perception is applied, in using mobile terminal The data putting sensor acquisition judge to identify the scene that user gathers decibel data, and this measurement data record is carried out scene Mark, thus speculate that user gathers the scene of data.

In order to achieve the above object, the present invention is achieved by the following technical solutions:

A kind of user's scene recognition method based on mobile terminal Noise map, comprises the following steps:

Step one: determine and obtain the built-in sensors data of mobile terminal.Application according to intelligent perception Noise map Feature, and the different characteristics that the kinestate of mobile terminal is monitored by various kinds of sensors, determine and use four class sensors, line Property acceleration transducer, GPS sensor, gyro sensor and Proximity Sensor.

Step 2: user gathers the classification of scene coarseness.Microphone sensor is higher to the sensitivity of environmental change, through reality Test analysis to learn, shake, rub, the position of the placement of mike all can produce large effect to experimental result, therefore this The bright scene that gathers user carries out coarseness and fine-grained division.

Step 3: obtain fine granularity scene sensor judgment threshold.The present invention use four class sensors carry out scene knowledge Not, linear acceleration sensors identification user's low speed moves or resting state, gyro sensor identification intelligent terminal rotating Angle, Proximity Sensor identifies whether to have physics to block or wrap up terminal, and GPS sensor identification user is at a high speed or low speed Mobile status.

Step 4: user gathers scene and speculates sorting algorithm.The present invention designs a kind of for noise decibel data collection Scene classification identification process, is refined as scene six big classes, and uses linear transducer, gyro sensor, electrical distance to sense Device, GPS sensor carry out different scene partitioning to mobile terminal physical state.

In above-mentioned steps one, determine and to obtain the built-in sensors data method of mobile terminal as follows:

The purposes of four class sensors is as follows: use linear acceleration sensors, it is thus achieved that the axial acceleration of X, Y, Z tri- Angle value (m/s²), the most do not include acceleration of gravity；Use the geographical location information of GPS sensor user in real, and count Calculation obtains user moving speed (m/s)；Use gyro sensor to obtain the axial angular velocity of X, Y, Z tri-, and be calculated Mobile phone anglec of rotation speed (°)；Proximity Sensor is used to obtain the object distance away from mobile phone.

In above-mentioned steps two, it is as follows that user gathers scene coarseness sorting technique:

It is first depending on the scene influence degree to mike decibel data, carries out the division of coarseness, and clearly in scene The kind of sensor used in identification, afterwards, gathers the feature of scene, scene is refined as according to coarseness scene according to user Six class scenes, six class scenes are:

Scene 1, mobile phone are static is placed in non-vibrations physical surface, and mobile phone faces up and do not has shelter, without physically encapsulation；

Scene 2, mobile phone are placed in user's hands and user is without kinestate, and mobile phone is without frequently significantly flip angle and vibrations Phenomenon；

Scene 3, mobile phone are placed in the static physics having parcel ability, without significantly upset, seismism

In scene 4, mobile phone are placed in the object that faster moves or surface, is in open environment without physically encapsulation, without bigger Vibrations

In scene 5, mobile phone are placed in the physics of relatively slower motion or surface, is in open environment without physically encapsulation, without bigger Vibrations

Scene 6, object are placed in the closed occupancy of moving movement, have parcel phenomenon；

The decibel quality of data gathering each scene is marked, when the foundation of scoring gathers for outer bound pair smart mobile phone Vibrations and the impact of friction, by experimental analysis, mobile phone is static or to be positioned at decibel quality of data time in user's hands higher.

In above-mentioned steps three, obtain fine granularity scene sensor judgment threshold method as follows:

For above four class sensor scene judgment thresholds, gradient descent method is used to obtain different sensors judgment threshold.

The first step, uses smart mobile phone many groups of hybrid measurement test data under six scenes, and uses semiautomatic fashion By artificial mark, the data of six scenes are carried out scene identity, for verifying the accuracy of later stage algorithm scene classification.

Second step, calculates the judgment threshold obtaining different sensors under different scenes.

For GPS longitude and latitude data, formula is used to be calculated A between 2

(lng1, lat1), distance S (km) of B (lng2, lat2), i.e.

$S=2\×\arcsin \left\{\sqrt{{\left(\sin \frac{a}{2}\right)}^2+\cos \left(lat1\right)\×\cos \left(lat2\right)\×{\left(\sin \frac{b}{2}\right)}^2}\right\}\×6378.137$

Wherein a=lat1-lat2, b=lng1-lng2,6378.137 (km) are earth radius, thus obtain user and move SpeedWherein S and interval is respectively distance and the time interval of point-to-point transmission；Linear acceleration sensing Data (acc_x, acc_y, acc_z), use resultant acceleration accValue and average energy accEnergy to calculate the movement of user State, i.e.

$accValue=\sqrt{acc\_x^2+acc\_y^2+acc\_z^2}$

$accEnergy=\underset{i=1}{\overset{N}{\Σ}}\frac{{\left({\mathrm{accValue}}_i\right)}^2}{N}$

Linear acceleration average energy is used to judge walking, static acceleration average energy threshold value.Use gradient decline side Method calculates linear acceleration threshold value, y=b (min <b < max), and wherein b represents the cutting horizontal line intercept in y-axis, min and max divides Do not represent the minimum and maximum value of average energy in testing data.UseSensor scene when calculating y=b Classification error rate, wherein S_AlwaysRepresent the data count of the different scenes using semi-artificial mark, S_CorrectlyRepresent and use acceleration energy The number that can correctly divide data scene obtained after threshold decision.In calculating threshold process, b initializes value max, it The rear continuous iteration of square b of cutting reduces, and when error rate Δ≤0.01, iterative process stops, and the b value now meeting condition is i.e. divided Not for walk and the judgment threshold of resting state.Wherein when Δ > 0.05 time, b reduces 0.1 every time；When Δ≤0.05, b subtracts every time Little by 0.01.Gyroscope measured value is used for distinguishing whether mobile phone remains static, and this value is used for distinguishing mobile phone and is positioned at resting state Or it is in user's hands.For Android closely induction apparatus, one has two class values, when this value is 0, shows there is thing Body blocks, and shows to block without physics when this value is not 0, and mobile phone is in open space.Gyroscope is used to distinguish whether mobile phone is located In relative static conditions, this value is used for distinguishing mobile phone and is positioned at resting state or is in user's hands.Through mixing scene not To calculating iteration, obtain sensor judgment threshold.Test result indicate that, when mobile terminal is in absolute rest state, gyro Instrument and accekeration are not 0, according to the difference of mobile phone hardware, have different deviants.

In above-mentioned steps four, user gathers scene and speculates that sorting algorithm method is as follows:

First each sensor judgment threshold is initialized；GPS is used to calculate user moving speed, if high-speed moving state v >=4.6m/s, it is judged that whether Proximity Sensor is 0, if not 0 shows unobstructed, is now labeled as scene 4；If low speed Mobile status v≤4.6m/s, it is judged that whether average energy exceedes walking threshold value, if exceeding, it is judged that whether its Proximity Sensor It is 0, if not 0, then it is marked as scene 5, if 0, then it is marked as scene 6；If average energy threshold value is not above Walking threshold value, then judge whether Proximity Sensor is 0, if 0, then it is marked as scene 3, if Proximity Sensor is not 0, it is judged that it meets average energy the most simultaneously and exceedes static average energy threshold value and the gyroscope anglec of rotation more than time static Gyroscope angle, if being unsatisfactory for, being labeled as scene 1, being otherwise marked as scene 2.

Beneficial effects of the present invention: user's scene Recognition algorithm based on mobile terminal Noise map that the present invention provides, Decibel data acquisition scenarios under the existing user's difference behavioral pattern of calculating and sending in real time, it is possible to effectively analytical calculation user participates in ring The quality of data of border perception, based on this by effectively filtering and data prediction, it is possible to be effectively improved the visualization of Noise map, And advisory opinion can be provided to user's incentive measure based on intelligent perception, thus more people is encouraged to participate in community or even city In the environmental monitoring task in city.

Accompanying drawing explanation

Fig. 1 is mobile terminal scene Recognition flow chart schematic diagram of the present invention.

Detailed description of the invention

The invention will be further described below in conjunction with the accompanying drawings.

A kind of user's scene recognition method based on mobile terminal Noise map, comprises the following steps:

Step one: determine and obtain the built-in sensors data of mobile terminal.Application according to intelligent perception Noise map Feature, and the different characteristics that the kinestate of mobile terminal is monitored by various kinds of sensors, determine and use four class sensors, line Property acceleration transducer, GPS sensor, gyro sensor and Proximity Sensor.

Step 2: user gathers the classification of scene coarseness.Microphone sensor is higher to the sensitivity of environmental change, through reality Test analysis to learn, shake, rub, the position of the placement of mike all can produce large effect to experimental result, therefore this The bright scene that gathers user carries out coarseness and fine-grained division.

Step 3: obtain fine granularity scene sensor judgment threshold.The present invention use four class sensors carry out scene knowledge Not, linear acceleration sensors identification user's low speed moves or resting state, gyro sensor identification intelligent terminal rotating Angle, Proximity Sensor identifies whether to have physics to block or wrap up terminal, and GPS sensor identification user is at a high speed or low speed Mobile status.

Step 4: user gathers scene and speculates sorting algorithm.The present invention designs a kind of for noise decibel data collection Scene classification identification process, is refined as scene six big classes, and uses linear transducer, gyro sensor, electrical distance to sense Device, GPS sensor carry out different scene partitioning to mobile terminal physical state.

In above-mentioned steps one, determine and to obtain the built-in sensors data method of mobile terminal as follows:

The purposes of four class sensors is as follows: use linear acceleration sensors, it is thus achieved that the axial acceleration of X, Y, Z tri- Angle value (m/s²), the most do not include acceleration of gravity；Use the geographical location information of GPS sensor user in real, and count Calculation obtains user moving speed (m/s)；Use gyro sensor to obtain the axial angular velocity of X, Y, Z tri-, and be calculated Mobile phone anglec of rotation speed (^o)；Proximity Sensor is used to obtain the object distance away from mobile phone.

In above-mentioned steps two, it is as follows that user gathers scene coarseness sorting technique:

It is first depending on the scene influence degree to mike decibel data, carries out the division of coarseness, and clearly in scene The kind of sensor used in identification, as shown in table 1,

Table 1 user's scene coarseness is classified

Afterwards, gather the feature of scene according to user, according to the coarseness scene of table 1, scene be refined as six class scenes, As shown in table 2,

Table 2 user gathers scene partition by fine granularities and decibel data standards of grading

The decibel quality of data gathering each scene is marked, when the foundation of scoring gathers for outer bound pair smart mobile phone Vibrations and the impact of friction, by experimental analysis, mobile phone is static or to be positioned at decibel quality of data time in user's hands higher.

In above-mentioned steps three, obtain fine granularity scene sensor judgment threshold method as follows:

For above four class sensor scene judgment thresholds, gradient descent method is used to obtain different sensors judgment threshold.

The first step, uses smart mobile phone many groups of hybrid measurement test data under six scenes, and uses semiautomatic fashion By artificial mark, the data of six scenes are carried out scene identity, for verifying the accuracy of later stage algorithm scene classification.

Second step, calculates the judgment threshold obtaining different sensors under different scenes.

For GPS longitude and latitude data, formula is used to be calculated A between 2

(lng1, lat1), distance S (km) of B (lng2, lat2), i.e.

$S=2\&times;\arcsin \left\{\sqrt{{\left(\sin \frac{a}{2}\right)}^2+\cos \left(lat1\right)\&times;\cos \left(lat2\right)\&times;{\left(\sin \frac{b}{2}\right)}^2}\right\}\&times;6378.137$

Wherein a=lat1-lat2, b=lng1-lng2,6378.137 (km) are earth radius, thus obtain user and move SpeedWherein S and interval is respectively distance and the time interval of point-to-point transmission；Linear acceleration sensing Data (acc_x, acc_y, acc_z), use resultant acceleration accValue and average energy accEnergy to calculate the movement of user State, i.e.

$accValue=\sqrt{acc\_x^2+acc\_y^2+acc\_z^2}$

$accEnergy=\underset{i=1}{\overset{N}{\&Sigma;}}\frac{{\left({\mathrm{accValue}}_i\right)}^2}{N}$

In inventive algorithm, linear acceleration average energy is used to judge walking, static acceleration average energy threshold value. Using gradient descent method to calculate linear acceleration threshold value, y=b (min <b < max), wherein b represents cutting horizontal line cutting in y-axis Away from, min and max represents the minimum and maximum value of average energy in testing data respectively.UseCalculate y=b Time sensor scene classification error rate, wherein S_AlwaysRepresent the data count of the different scenes using semi-artificial mark, S_CorrectlyExpression makes With the number that can correctly divide data scene obtained after acceleration energy threshold decision.In calculating threshold process, b is initial Changing value max, cut the continuous iteration of square b afterwards and reduce, when error rate Δ≤0.01, iterative process stops, and now meets The b value of condition is i.e. respectively the judgment threshold of walking and resting state.Wherein when Δ > 0.05 time, b reduces 0.1 every time；When Δ≤ When 0.05, b reduces 0.01 every time.In inventive algorithm, gyroscope measured value is used for distinguishing whether mobile phone remains static, This value is used for distinguishing mobile phone and is positioned at resting state or is in user's hands.For Android closely induction apparatus, one has two Class value, when this value is 0, shows have object to block, shows to block without physics when this value is not 0, and mobile phone is in open sky Between.Using gyroscope to distinguish whether mobile phone is in relative static conditions, this value is used for distinguishing mobile phone and is positioned at resting state or place In user's hands.Through mixing scene not to calculate iteration, the sensor judgment threshold that inventive algorithm obtains such as table 3 institute Show.

Table 3Android sensor scene judgment threshold

Test result indicate that, when mobile terminal is in absolute rest state, gyroscope and accekeration are not 0, root According to the difference of mobile phone hardware, there is different deviants.

In above-mentioned steps four, user gathers scene and speculates that sorting algorithm method is as follows:

In user gathers scene Recognition algorithm, scene Recognition flow process is as shown in Figure 1.First initialize each sensor to judge Threshold value；GPS is used to calculate user moving speed, if high-speed moving state v >=4.6m/s, it is judged that whether Proximity Sensor is 0, if not 0 shows unobstructed, now it is labeled as scene 4；If low mobility state v≤4.6m/s, it is judged that average energy Whether exceed walking threshold value, if exceeding, it is judged that whether its Proximity Sensor is 0, if not 0, then it is marked as scene 5, If 0, then it is marked as scene 6；If average energy threshold value is not above threshold value of walking, then whether judge Proximity Sensor It is 0, if 0, then it is marked as scene 3, if Proximity Sensor is not 0, it is judged that it meets average energy the most simultaneously and surpasses Crossing static average energy threshold value and the gyroscope anglec of rotation more than gyroscope angle time static, if being unsatisfactory for, being labeled as scene 1, otherwise it is marked as scene 2.

Claims (5)

1. user's scene recognition method based on mobile terminal Noise map, it is characterised in that: comprise the following steps:

Step one: determine and obtain the built-in sensors data of mobile terminal；

According to the application characteristic of intelligent perception Noise map, and the kinestate of mobile terminal is monitored not by various kinds of sensors Same feature, determines and uses four class sensors, linear acceleration sensors, GPS sensor, gyro sensor and closely Sensor；

Step 2: user gathers the classification of scene coarseness；

User is gathered scene and carries out coarseness and fine-grained division；

Step 3: obtain fine granularity scene sensor judgment threshold；

Described linear acceleration sensors identification user's low speed moves or resting state, described gyro sensor identification intelligent Terminal rotating angle, described Proximity Sensor identifies whether to have physics to block or wrap up terminal, described GPS sensor identification User's high speed or low mobility state；

Step 4: user gathers scene and speculates sorting algorithm；

Use the scene classification identification process for noise decibel data collection, scene is refined as six big classes, and uses linear Sensor, gyro sensor, Proximity Sensor, GPS sensor carry out different scene partitioning to mobile terminal physical state.

A kind of user's scene recognition method based on mobile terminal Noise map the most according to claim 1, its feature exists In: in described step one, the built-in sensors data method obtaining mobile terminal is as follows:

Use linear acceleration sensors, it is thus achieved that the axial accekeration of X, Y, Z tri-, the most do not include acceleration of gravity；

Use the geographical location information of GPS sensor user in real, and be calculated user moving speed；

Use gyro sensor to obtain the axial angular velocity of X, Y, Z tri-, and be calculated mobile phone anglec of rotation speed；

Proximity Sensor is used to obtain the object distance away from mobile phone.

A kind of user's scene recognition method based on mobile terminal Noise map the most according to claim 1, its feature exists In: in described step 2, it is as follows that user gathers scene coarseness sorting technique:

(3-1), it is first depending on the scene influence degree to mike decibel data, carries out the division of coarseness, and the most on the scene The kind of sensor used in scape identification；

(3-2), then, gather the feature of scene according to user, according to coarseness scene, scene is refined as six class scenes, described Six class scenes are:

Scene 1, mobile phone are static is placed in non-vibrations physical surface, and mobile phone faces up and do not has shelter, without physically encapsulation；

Scene 2, mobile phone are placed in user's hands and user is without kinestate, and mobile phone is existing without the most significantly flip angle and vibrations As；

Scene 3, mobile phone are placed in the static physics having parcel ability, without significantly upset, seismism

In scene 4, mobile phone are placed in the object that faster moves or surface, is in open environment without physically encapsulation, without significant shock

In scene 5, mobile phone are placed in the physics of relatively slower motion or surface, is in open environment without physically encapsulation, without significant shock

Scene 6, object are placed in the closed occupancy of moving movement, have parcel phenomenon；

(3-3) the decibel quality of data, to each scene gathered is marked, gathering according to for outer bound pair smart mobile phone of scoring Shi Zhendong and the impact of friction.

A kind of user's scene recognition method based on mobile terminal Noise map the most according to claim 1, its feature exists In: in described step 3, obtain fine granularity scene sensor judgment threshold method as follows:

For four class sensor scene judgment thresholds, gradient descent method is used to obtain different sensors judgment threshold；

(4-1), use smart mobile phone many groups of hybrid measurement test data under six scenes, and use semiautomatic fashion to pass through people Work mark carries out scene identity to the data of six scenes, for verifying the accuracy of later stage algorithm scene classification；

(4-2) judgment threshold obtaining different sensors under different scenes, is calculated；

GPS sensor:

For GPS longitude and latitude data, formula is used to be calculated the distance of A (lng1, lat1) between 2, B (lng2, lat2) S (km), i.e.

$S=2\&times;\arcsin \left\{\sqrt{{\left(\sin \frac{a}{2}\right)}^2+\cos \left(lat1\right)\&times;\cos \left(lat2\right)\&times;{\left(\sin \frac{b}{2}\right)}^2}\right\}\&times;6378.137$

Wherein a=lat1-lat2, b=lng1-lng2,6378.137 (km) are earth radius, thus obtain user moving speedWherein S and interval is respectively distance and the time interval of point-to-point transmission.

Linear acceleration sensors:

Linear acceleration sensing data (acc_x, acc_y, acc_z), uses resultant acceleration accValue and average energy AccEnergy calculates the mobile status of user, i.e.

$accValue=\sqrt{acc\_x^2+acc\_y^2+acc\_z^2}$

$accEnergy=\underset{i=1}{\overset{N}{\&Sigma;}}\frac{{\left({\mathrm{accValue}}_i\right)}^2}{N};$

Use linear acceleration average energy to judge walking, static acceleration average energy threshold value, use gradient descent method meter Calculate linear acceleration threshold value,

Y=b (min <b < max)

Wherein b represent cutting horizontal line in the intercept of y-axis, min and max represents the minimum and of average energy in testing data respectively Big value, usesSensor scene classification error rate when calculating y=b, wherein S_AlwaysRepresent and use semi-artificial mark The data count of different scenes, S_CorrectlyRepresent that obtain after using acceleration energy threshold decision can correctly divide data fields The number of scape；Calculating in threshold process, b initializes value max, cuts the reduction of square b continuous iteration afterwards, when error rate Δ≤ When 0.01, iterative process stops, and the b value now meeting condition is i.e. respectively the judgment threshold of walking and resting state, its In when Δ > 0.05 time, b reduces 0.1 every time；When Δ≤0.05, b reduces 0.01 every time；

Gyro sensor:

Gyroscope measured value is used for distinguishing whether mobile phone remains static, and this is used for distinguishing mobile phone and is positioned at resting state or place In user's hands；

For Android closely induction apparatus, one has two class values, when this value is 0, shows have object to block, at mobile phone In closed state；Showing to block without physics when this value is not 0, mobile phone is in open space；

Using gyroscope to distinguish whether mobile phone is in relative static conditions, this value is used for distinguishing mobile phone and is positioned at resting state or place In user's hands, through mixing scene not to calculate iteration, obtain sensor judgment threshold.

A kind of user's scene recognition method based on mobile terminal Noise map the most according to claim 1, its feature exists In: in described step 4, user gathers scene and speculates that sorting algorithm method is as follows:

First each sensor judgment threshold is initialized；GPS is used to calculate user moving speed, if high-speed moving state v >= 4.6m/s, it is judged that whether Proximity Sensor is 0, if not 0 shows unobstructed, is now labeled as scene 4；If low speed moves Dynamic state v≤4.6m/s, it is judged that whether average energy exceedes walking threshold value, if exceeding, it is judged that whether its Proximity Sensor is 0, if not 0, then it is marked as scene 5, if 0, then it is marked as scene 6；If average energy threshold value is not above row Walk threshold value, then judge whether Proximity Sensor is 0, if 0, then it is marked as scene 3, if Proximity Sensor is not 0, Judge that it meets average energy the most simultaneously and exceedes static average energy threshold value and the gyroscope anglec of rotation more than top time static Spiral shell instrument angle, if being unsatisfactory for, being labeled as scene 1, being otherwise marked as scene 2.