CN111329485A - Gait recognition method and device based on IMU - Google Patents

Gait recognition method and device based on IMU Download PDF

Info

Publication number
CN111329485A
CN111329485A CN202010117483.9A CN202010117483A CN111329485A CN 111329485 A CN111329485 A CN 111329485A CN 202010117483 A CN202010117483 A CN 202010117483A CN 111329485 A CN111329485 A CN 111329485A
Authority
CN
China
Prior art keywords
state
probability
gait
hmm
motion data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010117483.9A
Other languages
Chinese (zh)
Inventor
吴庆勋
崔翔
王道臣
刘昊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Machinery Equipment Research Institute
Original Assignee
Beijing Machinery Equipment Research Institute
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Machinery Equipment Research Institute filed Critical Beijing Machinery Equipment Research Institute
Priority to CN202010117483.9A priority Critical patent/CN111329485A/en
Publication of CN111329485A publication Critical patent/CN111329485A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/10—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration
    • G01C21/12—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning
    • G01C21/16—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 by using measurements of speed or acceleration executed aboard the object being navigated; Dead reckoning by integrating acceleration or speed, i.e. inertial navigation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
    • A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
    • A61B5/112—Gait analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Dentistry (AREA)
  • Medical Informatics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Physiology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Artificial Intelligence (AREA)
  • Probability & Statistics with Applications (AREA)
  • Automation & Control Theory (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
  • Image Analysis (AREA)

Abstract

The application discloses a gait recognition method and device based on an IMU (inertial measurement Unit), and belongs to the technical field of computers. The method comprises the steps of acquiring motion data during gait switching by utilizing an IMU (inertial measurement Unit), and preprocessing the motion data; extracting a motion characteristic vector at the t moment from the preprocessed motion data, and inputting the motion characteristic vector serving as an observation sequence into a pre-trained Hidden Markov Model (HMM); and calculating the HMM output probability and each state probability at the time t, and taking the gait state corresponding to the state with the maximum state probability as the gait state at the time t. According to the gait prediction method and device, the data acquired by the IMU are used as the observation value of the HMM, the gait phase is used as the state value of the HMM, the gait is predicted through the HMM algorithm, and the gait prediction accuracy is improved.

Description

Gait recognition method and device based on IMU
Technical Field
The invention belongs to the technical field of computers, and relates to a gait recognition method and device based on an IMU (inertial measurement Unit).
Background
The human body lower limb gait recognition problem for exoskeleton control is to judge the motion type of lower limbs and the phase under a certain motion in real time so as to realize the accuracy and real-time performance of a recognition result. For human walking gait, the walking of the lower limbs has rhythmic gait, the motion information of each part of the lower limbs shows periodic change, one gait cycle is divided into phase time sequences according to gait characteristics, for example, the phase of the walking cycle is sequentially circulated by five phases of tiptoe off, swing middle, heel landing, sole supporting and the like.
One common way to identify gait of lower limbs is by collecting angle information of each part of the lower limbs. Because the sensor measures noise randomly and is influenced by individual difference, walking characteristics, environmental conditions and the like, and the repeatability of the external motion posture characteristics of each part of the lower limb in each motion period is poor, the method of judging only by angle information is easy to cause misjudgment.
Disclosure of Invention
In order to solve the problem that misjudgment is easily caused when gait is judged only by adopting angle information in the related technology, the application provides a gait recognition method and a device based on IMU, and the technical scheme is as follows:
in a first aspect, an IMU-based gait recognition method is provided, which includes:
utilizing an IMU to acquire motion data during gait switching, and preprocessing the motion data;
extracting a motion characteristic vector at the t moment from the preprocessed motion data, and inputting the motion characteristic vector serving as an observation sequence into a pre-trained Hidden Markov Model (HMM);
and calculating the HMM output probability and each state probability at the time t, and taking the gait state corresponding to the state with the maximum state probability as the gait state at the time t.
Optionally, the obtaining, by the IMU, the motion data during gait switching, and preprocessing the motion data includes:
and obtaining motion data by utilizing the IMU, and smoothing, filtering and normalizing the motion data characteristics, wherein the motion data comprises the speed and the acceleration during gait switching.
Optionally, the extracting a motion feature vector at time t from the preprocessed motion data includes:
dividing the preprocessed motion data by adopting a sliding window;
and extracting a motion characteristic vector from the motion data in the t-th window as the motion characteristic vector at the t moment.
Optionally, the calculating the HMM output probability and each state probability at the time t includes:
and calculating the HMM output probability and each state probability at the t moment by utilizing the HMM output probability at the t-1 moment and a Viterbi algorithm.
Optionally, the method provided by the present application further includes:
taking the motion data in the training sample as an observation sequence, and taking the reference phase data in the training sample as a state sequence;
estimating an initial state probability distribution matrix pi, a state transition probability distribution matrix A and an observation probability distribution matrix B of the HMM according to the observation sequence and the state sequence, so that the probability of the observation sequence of the HMM under the parameters pi, A and B is maximum;
wherein, the element pi in the probability distribution matrix pi of the initial stateiIs the state i at the first moment in the state sequence1Is s isiProbability of time;
state transition probability distribution matrix
Figure BDA0002391946090000021
Middle element aijIs a state s at the time t-1iTransition to time t state sjThe probability of (A);
observing a probability distribution matrix
Figure BDA0002391946090000022
Middle element bjkIs in a state sjTime observation quantity otIs v iskThe probability of (d);
HMM has a phase state space of S ═ S1,s2,…sN-said state sequence is I ═ I1,i2,…iTAnd s.t.i ∈ S, HMM with an observation space of V ═ V1,v2,…vMAssuming a sequence length of T, the observed sequence is O ═ O1,o2,…oTH, and s.t.o ∈ V.
In a second aspect, an IMU-based gait recognition apparatus is provided, the apparatus comprising:
the acquisition module is configured to acquire motion data during gait switching by utilizing the IMU and preprocess the motion data;
the extraction module is configured to extract a motion characteristic vector at the time t from the motion data preprocessed by the acquisition module;
an input module configured to input the motion feature vector extracted by the extraction module as an observation sequence into a pre-trained Hidden Markov Model (HMM);
and the gait determining module is configured to use the HMM output probability and each state probability at the time t to calculate the gait state corresponding to the state with the maximum state probability as the gait state at the time t.
Optionally, the obtaining module is further configured to:
and obtaining motion data by utilizing the IMU, and smoothing, filtering and normalizing the motion data characteristics, wherein the motion data comprises the speed and the acceleration during gait switching.
Optionally, the extracting module includes:
a segmentation unit configured to segment the preprocessed motion data using a sliding window;
and the extracting unit is configured to extract the motion characteristic vector from the motion data in the t-th window divided by the dividing unit as the motion characteristic vector at the t moment.
Optionally, the gait determination module is further configured to:
and calculating the HMM output probability and each state probability at the t moment by utilizing the HMM output probability at the t-1 moment and a Viterbi algorithm.
Optionally, the apparatus provided in the present application further includes:
the training module is configured to take the motion data in the training sample as an observation sequence and take the reference phase data in the training sample as a state sequence; estimating an initial state probability distribution matrix pi, a state transition probability distribution matrix A and an observation probability distribution matrix B of the HMM according to the observation sequence and the state sequence determined by the determining module, so that the probability of the observation sequence of the HMM under the parameters pi, A and B is maximum;
wherein, the element pi in the probability distribution matrix pi of the initial stateiIs the state i at the first moment in the state sequence1Is s isiProbability of time;
state transition probability distribution matrix
Figure BDA0002391946090000031
Middle element aijIs a state s at the time t-1iTransition to time t state sjThe probability of (A);
observing a probability distribution matrix
Figure BDA0002391946090000032
Middle element bjkIs in a state sjTime observation quantity otIs v iskThe probability of (d);
HMM has a phase state space of S ═ S1,s2,…sN-said state sequence is I ═ I1,i2,…iTAnd s.t.i ∈ S, HMM with an observation space of V ═ V1,v2,…vMAssuming a sequence length of T, the observed sequence is O ═ O1,o2,…oTH, and s.t.o ∈ V.
Based on the technical characteristics, the application can at least realize the following technical effects:
the data acquired by the IMU is used as the observation value of the HMM, the gait phase is used as the state value of the HMM, the gait is predicted through the HMM algorithm, and the gait prediction accuracy is improved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.
FIG. 1 is a schematic illustration of motion data acquired by an IMU during a walking cycle as provided in one embodiment of the present application;
FIG. 2 is a flow chart of HMM training in an IMU based gait recognition method provided in one embodiment of the present application;
figure 3 is a flow chart of a method of IMU based gait recognition provided in one embodiment of the present application;
fig. 4A is a schematic diagram of an IMU-based gait recognition apparatus provided in an embodiment of the present application;
fig. 4B is a schematic structural diagram of an IMU-based gait recognition apparatus according to another embodiment of the present application.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the invention, as detailed in the appended claims.
The human lower limb gait recognition problem for exoskeleton control is to judge the motion type of the lower limb and the phase under a certain motion in real time, in order to realize the accuracy and real-time performance of a recognition result, the angle and acceleration characteristics of each part of the lower limb are combined, the angle characteristics can generally accurately correspond to the phase of a gait cycle, and the acceleration can be regarded as the trend of the rotary motion of the joint of the lower limb, namely as the leading index of the phase change. The Inertial Measurement Unit (IMU) is small and easy to use, can acquire rigid body kinematics information in real time, is fixedly connected with the IMU on the back, the thigh, the crus and the foot, and can acquire the speed and acceleration information of each part of the lower limb through rigid body motion transformation. The Hidden Markov Model (HMM) is a time sequence variability signal statistical model, can describe the dynamic process and adapt to the change of the dynamic process, takes the lower limb kinematics signal and the gait phase as the observed value and the state value of the model respectively, describes each gait phase as a random state, and adopts the hidden Markov model to carry out the gait recognition corresponding to the switching process of the gait phase. Gait prediction based on IMU is realized through data processing, feature extraction and HMM algorithm design.
For human walking gait, lower limb walking has rhythmic gait, the motion information of each part of the lower limb, namely speed and acceleration, presents periodic change, one gait cycle is divided into phase sequence according to gait characteristics, as shown in figure 1, the motion phase of the walking cycle, five phases of tiptoe off, swing middle, heel landing, sole support and the like circulate in sequence, the ordinate of the first-line curve on the upper surface in figure 1 is knee joint angle data, and the abscissa is gait cycle; the ordinate of the second row of curves is knee joint angular velocity data, and the abscissa is the gait cycle. However, the sensor measures noise randomly, and is influenced by individual differences, walking characteristics, environmental conditions and the like, and the repeatability of the external motion posture characteristics of each part of the lower limb in each motion period is poor, so that erroneous judgment is easily caused only by adopting the angle threshold judgment method. And judging the prior trend of the movement of each part of the lower limb by combining the acceleration information, and positioning the gait switching time point by combining the judgment of an angle threshold value. In order to more accurately extract the movement and acceleration information during gait switching, preprocessing operations such as smoothing, filtering and normalization are carried out on IMU signals, and smoothing and filtering windows which are as small as possible are used for processing in order to control data errors and data real-time performance. And then, segmenting data by adopting a sliding window, and extracting kinematic features from a single window for subsequent HMM algorithm input.
The IMU-based gait recognition method provided by the present application is exemplified below with reference to fig. 2 and 3.
Before predicting gait by using the HMM model and the motion data acquired by the IMU, training the HMM model is first required, and a training process is shown in fig. 2, which is a flowchart of HMM training in the IMU-based gait recognition method provided in an embodiment of the present application, and the training process includes the following steps:
step 201, using motion data in a training sample as an observation sequence, and using reference phase data in the training sample as a state sequence;
the phase change process in one gait cycle during walking is tiptoe off, swing middle period, heel landing, sole landing and sole supporting in sequence. When the human body is divided into N phases (N is 5 in fig. 1) in the normal walking gait, there are N states corresponding to the markov process, and the phase state space of the HMM is S { S ═ S1,s2,…sNThe state sequence is I ═ I1,i2,…iTAnd s.t.i ∈ S, HMM with an observation space of V ═ V1,v2,…vMAssuming the sequence length is T, the observed sequence is O ═ O1,o2,…oTH, and s.t.o ∈ V.
Step 202, according to the observation sequence and the state sequence, estimating an initial state probability distribution matrix pi, a state transition probability distribution matrix A and an observation probability distribution matrix B of the HMM, so that the probability of the observation sequence of the HMM under the parameters pi, A and B is maximum.
The hidden Markov model has three elements: the initial state probability distribution matrix pi, the state transition probability distribution matrix A and the observation probability distribution matrix B are as follows:
element pi in initial state probability distribution matrix piiIs the state i at the first moment in the state sequence1Is s isiProbability of time, i.e. pi ═ pii＝P(i1＝si)}。
State transition probability distribution matrix
Figure BDA0002391946090000051
Middle element aijIs a state s at the time t-1iTransition to time t state sjProbability of (a)ij＝P(it＝sj|it-1＝si). Only one state can be in at a certain moment t, and the state i of the Markov chain at the momenttIs dependent on the state it-1And a state transition probability matrix a.
Observing a probability distribution matrix
Figure BDA0002391946090000052
Middle element bjkIs in a state sjTime observation quantity otIs v iskProbability of (i) bj(k)＝P(ot＝vk|it＝sj)。
And (3) applying a classical B-W algorithm, and adopting maximum likelihood estimation based on a recursive method to maximize the observation probability so as to obtain an HMM model under model parameters pi and A, B.
After training of the HMM model is completed, a gait is predicted by using the HMM model and motion data information acquired by the IMU, and a flow is shown in fig. 3, which is a flow chart of a gait recognition method based on the IMU provided in an embodiment of the present application, and the method includes the following steps:
step 301, obtaining motion data during gait switching by using an IMU (inertial measurement unit), and preprocessing the motion data;
the method comprises the steps of utilizing an IMU to obtain motion data during gait switching, utilizing the IMU to obtain the motion data during preprocessing of the motion data, and conducting smoothing, filtering and normalization processing on motion data characteristics, wherein the motion data comprises speed and acceleration during gait switching.
Step 302, extracting a motion characteristic vector at the time t from the preprocessed motion data, and inputting the motion characteristic vector into a pre-trained Hidden Markov Model (HMM) as an observation sequence;
the motion characteristic vector at the time t is extracted from the preprocessed motion data, firstly, the preprocessed motion data is divided by adopting a sliding window, and in practical application, in order to control data error and data real-time performance, a smoothing and filtering window which is as small as possible is used for processing.
Then, a motion feature vector is extracted from the motion data in the t-th window as a motion feature vector at time t.
Step 303, calculating HMM output probability and each state probability at time t, and setting the gait state corresponding to the state with the highest state probability as the gait state at time t.
Calculating the HMM output probability and each state probability at the time t, wherein the method comprises the following steps:
and calculating the HMM output probability and each state probability at the time t by using the HMM output probability at the time t-1 and a Viterbi algorithm, wherein the gait state corresponding to the state with the highest state probability is the gait state at the time t, and the phase identification of the lower limb movement gait is completed.
In summary, the method provided by the present application may include a training part and a gait recognition part:
a training part: and training the model to obtain HMM model parameters pi and A, B. Known observation sequence O ═ O1,o2,…oTAnd the state sequence I ═ I1,i2,…iTAnd estimating the model parameters to maximize the probability of the observation sequence under the model parameters. And (3) applying a classical B-W algorithm, and adopting maximum likelihood estimation based on a recursive method to maximize the observation probability so as to obtain an HMM model under model parameters pi and A, B.
A gait recognition part: and inputting the observation characteristic vector into a trained HMM model, and resolving a lower limb motion state value sequence, namely a gait phase. The method comprises the steps of acquiring human body lower limb kinematics information by utilizing an IMU, taking an extracted motion feature vector sequence as an input value of an HMM, extracting various data features at a moment t in an HMM gait state recognition stage, inputting the data features into a trained HMM, calculating the HMM output probability and each state probability by utilizing a Viterbi algorithm in combination with the HMM output probability at the previous moment, and completing the phase recognition of lower limb movement gait.
In summary, according to the gait recognition method based on the IMU, the data acquired by the IMU are used as the observation value of the HMM, the gait phase is used as the state value of the HMM, and the HMM algorithm is used to realize the gait prediction, so that the gait prediction accuracy is improved.
The technical scheme provided by the application starts from processing and identifying the motion signal, and improves the representation capability and accuracy of the signal to human gait.
Obviously, the gait recognition algorithm related to the present application is not limited to the walking gait recognition exemplified by applying the present invention, but can also be applied to other gaits, such as squat, climbing up and down steps, climbing up and down slopes, etc.
The following are embodiments of the disclosed apparatus that may be used to perform embodiments of the disclosed methods. For details not disclosed in the embodiments of the apparatus of the present disclosure, refer to the embodiments of the method of the present disclosure.
Fig. 4A is a schematic structural diagram of an IMU-based gait recognition apparatus provided in an embodiment of the present application, which may implement the IMU-based gait recognition methods shown in fig. 2 and 3 by software, hardware or a combination of software and hardware. The IMU-based gait recognition apparatus provided by the present application may include: an acquisition module 410, an extraction module 420, an input module 430, and a gait determination module 440.
The acquisition module 410 may be configured to pre-process the motion data by acquiring the motion data at gait switch by the IMU.
The extraction module 420 may be configured to extract a motion feature vector at time t from the motion data preprocessed by the acquisition module 410.
The input module 430 may be configured to input the motion feature vectors extracted by the extraction module 420 as an observation sequence into a pre-trained hidden markov model HMM.
The gait determination module 440 may be configured to use the HMM output probability and each state probability at the time t to calculate the gait state corresponding to the state with the highest state probability as the gait state at the time t.
In one possible implementation, the obtaining module 410 is further configured to obtain motion data by using the IMU, and perform smoothing, filtering, and normalization processing on the motion data, where the motion data includes velocity and acceleration at gait switching.
In another possible implementation manner, please refer to fig. 4B, which is a schematic structural diagram of an IMU-based gait recognition apparatus provided in another embodiment of the present application, where the extracting module 420 may include: a segmentation unit 421 and an extraction unit 422.
The segmentation unit 421 may be configured to segment the pre-processed motion data using a sliding window.
The extracting unit 422 may be configured to extract a motion feature vector from the motion data within the t-th window divided by the dividing unit 421 as a motion feature vector at time t.
In another possible implementation, the gait determination module 440 is further configured to: and calculating the HMM output probability and each state probability at the t moment by utilizing the HMM output probability at the t-1 moment and a Viterbi algorithm.
In another possible implementation, the apparatus provided herein may further include a training module 450.
The training module 450 may be configured to estimate an initial state probability distribution matrix pi, a state transition probability distribution matrix a, and an observation probability distribution matrix B of the HMM according to the observation sequence and the state sequence determined by the determining module, taking the motion data in the training sample as the observation sequence, taking the reference phase data in the training sample as the state sequence, so that the probability of the observation sequence of the HMM under the parameters pi, a, and B is maximum;
wherein, the element pi in the probability distribution matrix pi of the initial stateiIs the state i at the first moment in the state sequence1Is s isiProbability of time;
state transition probability distribution matrix
Figure BDA0002391946090000071
Middle element aijIs a state s at the time t-1iTransition to time t state sjThe probability of (A);
observing a probability distribution matrix
Figure BDA0002391946090000072
Middle element bjkTo be atState sjTime observation quantity otIs v iskThe probability of (d);
HMM has a phase state space of S ═ S1,s2,…sNThe state sequence is I ═ I1,i2,…iTAnd s.t.i ∈ S, HMM with an observation space of V ═ V1,v2,…vMAssuming the sequence length is T, the observed sequence is O ═ O1,o2,…oTH, and s.t.o ∈ V.
In summary, the gait recognition device based on the IMU provided by the application uses the data acquired by the IMU as the observed value of the HMM, uses the gait phase as the state value of the HMM, and realizes the gait prediction by the HMM algorithm, thereby improving the accuracy of the gait prediction.
Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.
It will be understood that the invention is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the invention is limited only by the appended claims.

Claims (10)

1. An IMU-based gait recognition method, the method comprising:
utilizing an IMU to acquire motion data during gait switching, and preprocessing the motion data;
extracting a motion characteristic vector at the t moment from the preprocessed motion data, and inputting the motion characteristic vector serving as an observation sequence into a pre-trained Hidden Markov Model (HMM);
and calculating the HMM output probability and each state probability at the time t, and taking the gait state corresponding to the state with the maximum state probability as the gait state at the time t.
2. The method of claim 1, wherein the obtaining of the motion data at gait switch by the IMU, and the preprocessing of the motion data, comprises:
and obtaining motion data by utilizing the IMU, and smoothing, filtering and normalizing the motion data characteristics, wherein the motion data comprises the speed and the acceleration during gait switching.
3. The method of claim 1, wherein the extracting the motion feature vector at the time t from the pre-processed motion data comprises:
dividing the preprocessed motion data by adopting a sliding window;
and extracting a motion characteristic vector from the motion data in the t-th window as the motion characteristic vector at the t moment.
4. The method according to claim 1, wherein the calculating the HMM output probability and each state probability at time t comprises:
and calculating the HMM output probability and each state probability at the t moment by utilizing the HMM output probability at the t-1 moment and a Viterbi algorithm.
5. The method according to any one of claims 1-4, further comprising:
taking the motion data in the training sample as an observation sequence, and taking the reference phase data in the training sample as a state sequence;
estimating an initial state probability distribution matrix pi, a state transition probability distribution matrix A and an observation probability distribution matrix B of the HMM according to the observation sequence and the state sequence, so that the probability of the observation sequence of the HMM under the parameters pi, A and B is maximum;
wherein, the element pi in the probability distribution matrix pi of the initial stateiIs shaped likeState i at the first time in the state sequence1Is s isiProbability of time;
state transition probability distribution matrix
Figure FDA0002391946080000011
Middle element aijIs a state s at the time t-1iTransition to time t state sjThe probability of (A);
observing a probability distribution matrix
Figure FDA0002391946080000012
Middle element bjkIs in a state sjTime observation quantity otIs v iskThe probability of (d);
HMM has a phase state space of S ═ S1,s2,…sN-said state sequence is I ═ I1,i2,…iTAnd s.t.i ∈ S, HMM with an observation space of V ═ V1,v2,…vMAssuming a sequence length of T, the observed sequence is O ═ O1,o2,…oTH, and s.t.o ∈ V.
6. An IMU-based gait recognition apparatus, characterized in that the apparatus comprises:
the acquisition module is configured to acquire motion data during gait switching by utilizing the IMU and preprocess the motion data;
the extraction module is configured to extract a motion characteristic vector at the time t from the motion data preprocessed by the acquisition module;
an input module configured to input the motion feature vector extracted by the extraction module as an observation sequence into a pre-trained Hidden Markov Model (HMM);
and the gait determining module is configured to use the HMM output probability and each state probability at the time t to calculate the gait state corresponding to the state with the maximum state probability as the gait state at the time t.
7. The apparatus of claim 6, wherein the acquisition module is further configured to:
and obtaining motion data by utilizing the IMU, and smoothing, filtering and normalizing the motion data characteristics, wherein the motion data comprises the speed and the acceleration during gait switching.
8. The apparatus of claim 6, wherein the extraction module comprises:
a segmentation unit configured to segment the preprocessed motion data using a sliding window;
and the extracting unit is configured to extract the motion characteristic vector from the motion data in the t-th window divided by the dividing unit as the motion characteristic vector at the t moment.
9. The apparatus of claim 6, wherein the gait determination module is further configured to:
and calculating the HMM output probability and each state probability at the t moment by utilizing the HMM output probability at the t-1 moment and a Viterbi algorithm.
10. The apparatus according to any one of claims 6-9, wherein the apparatus further comprises:
the training module is configured to use the motion data in the training sample as an observation sequence, use the reference phase data in the training sample as a state sequence, and estimate an initial state probability distribution matrix pi, a state transition probability distribution matrix A and an observation probability distribution matrix B of the HMM according to the observation sequence and the state sequence determined by the determining module, so that the probability of the observation sequence of the HMM under the parameters pi, A and B is maximum;
wherein, the element pi in the probability distribution matrix pi of the initial stateiIs the state i at the first moment in the state sequence1Is s isiProbability of time;
state transition probability distribution matrix
Figure FDA0002391946080000021
Middle element aijIs a state s at the time t-1iTransition to time t state sjThe probability of (A);
observing a probability distribution matrix
Figure FDA0002391946080000022
Middle element bjkIs in a state sjTime observation quantity otIs v iskThe probability of (d);
HMM has a phase state space of S ═ S1,s2,…sN-said state sequence is I ═ I1,i2,…iTAnd s.t.i ∈ S, HMM with an observation space of V ═ V1,v2,…vMAssuming a sequence length of T, the observed sequence is O ═ O1,o2,…oTH, and s.t.o ∈ V.
CN202010117483.9A 2020-02-25 2020-02-25 Gait recognition method and device based on IMU Pending CN111329485A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010117483.9A CN111329485A (en) 2020-02-25 2020-02-25 Gait recognition method and device based on IMU

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010117483.9A CN111329485A (en) 2020-02-25 2020-02-25 Gait recognition method and device based on IMU

Publications (1)

Publication Number Publication Date
CN111329485A true CN111329485A (en) 2020-06-26

Family

ID=71173967

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010117483.9A Pending CN111329485A (en) 2020-02-25 2020-02-25 Gait recognition method and device based on IMU

Country Status (1)

Country Link
CN (1) CN111329485A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113146611A (en) * 2020-12-29 2021-07-23 武汉理工大学 Rigid-flexible coupling exoskeleton robot motion mode identification method
CN115239767A (en) * 2022-09-22 2022-10-25 北京工业大学 Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150161516A1 (en) * 2013-12-06 2015-06-11 President And Fellows Of Harvard College Method and apparatus for detecting mode of motion with principal component analysis and hidden markov model
CN109492703A (en) * 2018-11-23 2019-03-19 河北工程大学 A kind of recognition methods of gait, system and terminal device

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150161516A1 (en) * 2013-12-06 2015-06-11 President And Fellows Of Harvard College Method and apparatus for detecting mode of motion with principal component analysis and hidden markov model
CN109492703A (en) * 2018-11-23 2019-03-19 河北工程大学 A kind of recognition methods of gait, system and terminal device

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
刘畅等: "基于隐马尔可夫模型的步态识别算法", 《计算机工程与设计》 *
张向刚等: "一种基于隐马尔科夫模型的步态识别算法", 《计算机科学》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113146611A (en) * 2020-12-29 2021-07-23 武汉理工大学 Rigid-flexible coupling exoskeleton robot motion mode identification method
CN113146611B (en) * 2020-12-29 2022-06-03 武汉理工大学 Rigid-flexible coupling exoskeleton robot motion mode identification method
CN115239767A (en) * 2022-09-22 2022-10-25 北京工业大学 Dynamic passenger flow behavior situation prediction method, system, storage medium and equipment

Similar Documents

Publication Publication Date Title
Zhang et al. Human activity recognition based on motion sensor using u-net
CN108256433B (en) Motion attitude assessment method and system
KR101815975B1 (en) Apparatus and Method for Detecting Object Pose
KR101549645B1 (en) Method and apparatus of recognizing facial expression using motion dictionary
CN109753891A (en) Football player's orientation calibration method and system based on human body critical point detection
KR20160031246A (en) Method and apparatus for gait task recognition
CN109948741A (en) A kind of transfer learning method and device
CN110738154A (en) pedestrian falling detection method based on human body posture estimation
CN111329485A (en) Gait recognition method and device based on IMU
CN109447128B (en) Micro-inertia technology-based walking and stepping in-place movement classification method and system
JP5604249B2 (en) Human body posture estimation device, human body posture estimation method, and computer program
CN113516063B (en) Motion pattern recognition method based on K-Means and gait cycle similarity
Wang et al. A2dio: Attention-driven deep inertial odometry for pedestrian localization based on 6d imu
CN114267088B (en) Gait information processing method and device and electronic equipment
CN115346272A (en) Real-time tumble detection method based on depth image sequence
Ning et al. Articulated model based people tracking using motion models
Guenterberg et al. Distributed continuous action recognition using a hidden markov model in body sensor networks
CN111104902A (en) Hemiplegia gait classification method based on graph convolution network
Gaud et al. Human gait analysis and activity recognition: A review
Balti et al. Gait analysis and detection of human pose diseases
CN112293295B (en) Milk cow early lameness recognition method and device
CN110163142A (en) Real-time gesture recognition method and system
Li et al. An automatic framework for generating Labanotation scores from continuous motion capture data
CN115105057A (en) Gait phase dividing and identifying method and system for transverse walking
Zhang et al. PCA & HMM based arm gesture recognition using inertial measurement unit

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination