CN107157450B - Quantitative assessment method and system for hand motion ability of Parkinson's disease people - Google Patents
Quantitative assessment method and system for hand motion ability of Parkinson's disease people Download PDFInfo
- Publication number
- CN107157450B CN107157450B CN201710463520.XA CN201710463520A CN107157450B CN 107157450 B CN107157450 B CN 107157450B CN 201710463520 A CN201710463520 A CN 201710463520A CN 107157450 B CN107157450 B CN 107157450B
- Authority
- CN
- China
- Prior art keywords
- data set
- training data
- classifier
- features
- feature
- 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.)
- Active
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/48—Other medical applications
- A61B5/4848—Monitoring or testing the effects of treatment, e.g. of medication
-
- 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
- A61B5/1124—Determining motor skills
- A61B5/1125—Grasping motions of hands
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/389—Electromyography [EMG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
- A61B5/4082—Diagnosing or monitoring movement diseases, e.g. Parkinson, Huntington or Tourette
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Veterinary Medicine (AREA)
- Animal Behavior & Ethology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Artificial Intelligence (AREA)
- Pathology (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physiology (AREA)
- Neurology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Neurosurgery (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Developmental Disabilities (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Mathematical Physics (AREA)
- Fuzzy Systems (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Dentistry (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention provides a method and a system for quantitatively evaluating the hand motion ability of a Parkinson patient, wherein a wearable electromyographic sensor is used for capturing a hand surface electromyographic signal of a detected person when the detected person performs a specified action, and a trained motion ability classifier corresponding to the specified action is used for quantitatively evaluating the hand motion ability of the detected person by extracting time-domain features and frequency-domain features and features related to the completion condition of the specified action from the surface electromyographic signal. The method and the system can be used for more objectively and accurately evaluating the hand motion ability of the Parkinson patients.
Description
Technical Field
The invention relates to quantitative assessment of motion ability, in particular to a quantitative assessment method and a quantitative assessment system for hand motion ability of a Parkinson patient.
Background
The motor symptoms of patients with Parkinson's Disease (PD) are mainly manifested by bradykinesia, resting tremor, muscular rigidity and the like. Currently, the detection and assessment of the motor ability of PD patients mainly rely on the PD Rating Scale (UPDRS) provided by international motion society for grading, for example, the hand motor ability of parkinson patients is generally graded to 0-4, wherein 0 is equivalent to the hand motor ability of normal people, and the higher the grade, the worse the hand motor ability. However, such assessment is usually affected by the operation experience of the scoring physician and the state and emotion of the subject under evaluation, and thus the evaluation result is not sufficiently objective and accurate.
Disclosure of Invention
Therefore, the present invention is directed to overcoming the above-mentioned drawbacks of the prior art and providing a method and system for quantitatively evaluating the hand motor ability of a parkinson's disease person by using the surface electromyography of the hand.
The purpose of the invention is realized by the following technical scheme:
in one aspect, the invention provides a method for quantitatively evaluating hand motion capability of a person with parkinson's disease, comprising the following steps:
step 1, acquiring a surface electromyographic signal of a detected person when the detected person performs a specified action through an electromyographic sensor placed on a hand of the detected person;
the features used for training the parkinsonian movement ability classifier corresponding to the designated action comprise time domain features and frequency domain features extracted from the surface electromyographic signals and features which are acquired based on the surface electromyographic signals and are related to the completion condition of the designated action.
In the above method, the characteristic related to the completion of the specified action may include at least one of: the maximum electromyographic signal when the designated action is finished, and the time taken for finishing the designated action.
In the above method, the specified action may include at least one of: fist making and finger pointing.
The method may further comprise the step of training a parkinsonian motion capability classifier corresponding to the designated action, and the method comprises the following steps:
a) receiving a plurality of surface electromyographic signals which are acquired by electromyographic sensors worn by a plurality of Parkinson patients and normal persons with different degrees when the Parkinson patients and the normal persons perform the specified actions as training data sets;
b) extracting time domain characteristics and frequency domain characteristics from each surface electromyographic signal and acquiring characteristics related to the completion condition of the specified action based on the surface electromyographic signal;
c) selecting features for training the athletic performance classifier based on an information gain of each feature to a training data set, wherein the information gain of each feature to the training data set is a difference between an empirical entropy of the training data set and an empirical conditional entropy of the training data set given the feature;
d) training the athletic performance classifier based on the selected features.
In the above method, when the designated action is designated as the opposite finger, a sequence minimum optimization model may be used as the motion capability classifier; when the designated action is a punch, a J48 classifier may be employed as the athletic performance classifier.
In the above method, the empirical entropy of the training data set represents the uncertainty of classifying the training data set, and may be calculated by the following formula:
where D represents the training data set, H (D) represents the empirical entropy of the training data set D, n represents the total classification of the training data set D into n classes, piRepresenting the probability of classifying the data into class i.
In the above method, the empirical condition entropy of the training data set given a certain feature represents the uncertainty of classifying the training data set given the feature, and can be calculated by the following formula:
where A denotes a certain feature, H (D | A) is the empirical entropy of the training data set D given the feature A, pi(D | A) represents the probability of classifying the data into class i given the feature A.
In still another aspect, the present invention provides a system for quantitatively evaluating hand motion ability of a parkinson's disease person, comprising:
the device comprises a collecting device, a hand-operated sensor and a hand-operated control device, wherein the collecting device is used for collecting a surface electromyographic signal of a detected person when the detected person performs a specified action through an electromyographic sensor placed on the hand of the detected person;
the detection device is used for judging the grade of the hand movement ability of the detected person according to the collected surface electromyographic signals by utilizing a pre-trained Parkinson patient movement ability classifier corresponding to the specified action;
the features used for training the parkinsonian movement ability classifier corresponding to the designated action comprise time domain features and frequency domain features extracted from the surface electromyographic signals and features which are acquired based on the surface electromyographic signals and are related to the completion condition of the designated action.
In the above system, the characteristic related to the completion of the specified action may include at least one of: the maximum electromyographic signal when the designated action is finished, and the time taken for finishing the designated action.
In the above system, the specified action may include at least one of: fist making and finger pointing.
The system may further comprise training means for training a parkinsonian motor ability classifier corresponding to the specified action, configured to:
a) receiving a plurality of surface electromyographic signals which are acquired by electromyographic sensors worn by a plurality of Parkinson patients and normal persons with different degrees when the Parkinson patients and the normal persons perform the specified actions as training data sets;
b) extracting time domain characteristics and frequency domain characteristics from each surface electromyographic signal and acquiring characteristics related to the completion condition of the specified action based on the surface electromyographic signal;
c) selecting features for training the athletic performance classifier based on an information gain of each feature to a training data set, wherein the information gain of each feature to the training data set is a difference between an empirical entropy of the training data set and an empirical conditional entropy of the training data set given the feature;
d) training the athletic performance classifier based on the selected features.
In the above system, when the designated action is designated as the opposite finger, a sequence minimum optimization model can be used as the motion capability classifier; when the designated action is a punch, a J48 classifier may be employed as the athletic performance classifier.
Compared with the prior art, the invention has the advantages that:
the wearable electromyographic sensor is used for capturing the surface electromyographic signals of the hands of the detected person, so that the time domain characteristics and the frequency domain characteristics and the characteristics related to the completion condition of the specified action are extracted from the surface electromyographic signals to quantitatively evaluate the hand movement ability of the detected person, and the grade evaluation can be carried out on the hand movement ability of the Parkinson patients more objectively and accurately.
Drawings
Embodiments of the invention are further described below with reference to the accompanying drawings, in which:
FIG. 1 is a flow chart of a method for quantitatively evaluating hand motion ability of a Parkinson's disease person according to an embodiment of the invention;
FIG. 2(a) is a schematic diagram showing comparison of recognition accuracy of each classifier corresponding to the "finger-to-finger" operation;
FIG. 2(b) is a schematic diagram showing comparison of recognition accuracy of each classifier corresponding to the "fist making" action;
FIG. 3(a) is a diagram illustrating SMO classifier recognition performance trained using signal recognition features of surface electromyography signals for a "finger-to-finger" motion;
FIG. 3(b) is a schematic diagram of the J48 classifier identification performance trained by the signal identification characteristics of the surface electromyography signals for the "fist making" motion;
FIG. 4(a) is a diagram illustrating the performance of classifier identification corresponding to a "finger-to-finger" action trained using the method according to the present invention;
FIG. 4(b) is a schematic diagram of classifier identification performance corresponding to a "fist making" motion trained using the method according to the present invention;
FIG. 5(a) is a schematic diagram comparing recognition results of classifiers trained with different features and corresponding to the "finger-to-finger" action;
FIG. 5(b) is a schematic diagram comparing recognition results of classifiers corresponding to "fist making" actions trained by using different features.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail by embodiments with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The "finger-to-finger" and "fist-to-fist" gestures are two common international standard indicators used to assess hand motion ability in parkinson's patients. Hereinafter, how to evaluate the hand motion ability of the parkinson's disease person will be described with reference to specific embodiments by taking these two specific actions as examples. It is to be understood that embodiments of the present invention are not limited to specific athletic activities, and that the principles of the present invention may be applied to athletic performance assessment of other activities or other locations.
Fig. 1 shows a method for quantitatively evaluating hand motion ability of a parkinson's disease person according to an embodiment of the present invention, which includes acquiring a surface electromyographic signal of a subject performing a designated action via an electromyographic sensor placed at a relevant portion of the subject's hand (step 1); and judging the grade of the hand movement ability of the detected person according to the collected surface electromyographic signals by utilizing a pre-trained Parkinson patient movement ability classifier corresponding to the specified action (step 2). Wherein the parkinsonian motor ability classifier corresponding to a specified action is trained on the basis of sample data extracted when a large number of parkinsonian persons of different degrees and some normal persons perform the specified action.
More specifically, in step 1, the electromyographic sensor is placed at a designated location of the subject, such as an associated part of an arm, and the subject is then allowed to perform a designated action, such as making a fist or pointing. Here, any type of electromyographic sensor capable of acquiring a surface electromyographic signal of a human body may be employed. Preferably, a wearable electromyograph, such as an arm ring MYO manufactured by ThalmicLabs of canadian startup company, which is a gesture control arm ring, can be used to detect the surface electromyographic signals during hand movement, and can be wirelessly connected with other electronic products and transmit data through bluetooth. When a person to be detected makes a fist and acts on fingers, muscles above the lower arm mainly move to generate corresponding myoelectric signals, so that the MYO can be placed above the lower arm of the person to be detected. The electromyographic sensor sends the collected surface electromyographic signals to a corresponding signal analyzing and processing device, such as a mobile terminal such as a tablet or a mobile phone, a desktop computer, or any other computing device capable of processing the surface electromyographic signals, by wired or wireless transmission.
Preferably, the computing device responsible for processing the surface electromyographic signals may also perform a certain degree of preprocessing on the received surface electromyographic signals, such as filtering out noise or performing dimension reduction processing, such as synthesizing the signals from multiple electromyographic sensors, so as to simplify the computational complexity as much as possible and reduce the consumption of computational resources. Taking MYO as an example, the myoelectric muscle sensor comprises eight sensor chips, can acquire myoelectric signals of eight dimensions in one circle of the lower arm of a patient, and uses a for each myoelectric signal1,a2,a3,a4,a5,a6,a7,a8To indicate. When the detected person makes a fist and makes a finger, the muscle on the surface of the forearm is activated. For example, when a patient makes a fist, muscles around the lower arm are in a tight state at the same time, namely in an activated state, so that electromyographic signals of eight electromyographic sensors on the electromyograph are increased at the same time. When the arm of the patient is opened, the muscles on the upper circle of the forearm are in a relaxed state or an inactivated state at the same time, so that the myoelectric signals of the eight myoelectric sensors are reduced or disappeared at the same time. The collected eight-dimensional electromyographic signals have a correlation, so that the collected electromyographic signals can be synthesized by adopting an acceleration synthesis method, such as formula (1)Thereby obtaining a synthesized electromyographic signal sample. The synthesized electromyographic signals are used, so that the calculation complexity is reduced, the wearing direction of the electromyograph can be ignored by a wearer, and the electromyograph consisting of the eight electromyographic patches can be worn at will. It will be appreciated that the above pre-processing is performed in a preferred manner and not in any way limiting thereto, and that the use of surface electromyographic signals acquired by a single electromyographic sensor also allows a quantitative assessment of the hand's motor ability. The surface electromyographic signals mentioned below are no longer specifically electromyographic signals acquired by a single electromyographic sensor or synthesizedThe latter electromyographic signals, since they are processed in a similar manner.
With continued reference to fig. 1, after the surface electromyogram signal at the time of performing the specified action is obtained, the features required by the parkinsonian motor ability classifier corresponding to the specified action are extracted therefrom, and then the extracted features are provided to a classifier trained in advance to determine the level to which the hand motor ability of the subject belongs, in step 2. For each designated action, there is a pre-trained parkinsonian motor ability classifier corresponding thereto, and for the methods described below, the classifier corresponding to the fist making action is simply referred to as the fist making classifier, and the classifier corresponding to the finger action is simply referred to as the finger classifier. The training of the two classifiers is similar, and therefore, the training method of the classifier is only exemplified by the fist classifier.
During training, a large number of Parkinson patients and some normal people with different degrees wear the electromyographic sensors and perform a fist making action, so that a large number of surface electromyographic signals are collected to serve as a basic training data set. The key in training a classifier is to determine which features of the data are used for training, and these features directly affect the accuracy of the classification result. In general, a time domain and frequency domain analysis may be performed on the surface myoelectrical signal to extract a plurality of time domain features and frequency domain features therefrom to train a classifier. These time and frequency domain features reflect the characteristics of the electromyographic signal waveform itself. In a preferred embodiment of the invention, in addition to using time domain features and frequency domain features, features are employed that relate to the completion of a given action. This is because the characteristics related to the completion of a given action can be better differentiated with respect to the hand motion ability, considering that PD patients of different degrees have large differences in the time, strength, etc. used to complete a given action. In order to improve the accuracy of the classifier itself, the classifier is trained in the embodiment of the present invention by selecting the optimal features from the above-mentioned features based on the information gain. The process of training the classifier mainly comprises the following steps:
(1) extracting time domain features and frequency domain features from surface electromyogram signals
Surface electromyogram signal (SEMS) is a bioelectric signal emitted when neuromuscular activity is recorded from the surface of human skin through a sensor or an electrode, and is non-invasive. The amplitude of the surface electromyographic signal is stochastic and can be generally expressed as a quasi-gaussian distribution function. Generally, digital analysis is the main means for processing the surface electromyogram signal, and comprises time domain analysis and frequency domain analysis. The time domain analysis is to take the surface electromyogram signal as a function of time, and obtain some statistical characteristics of the electromyogram signal through analysis, for example, common time domain characteristics include an average value of the electromyogram signal in a time domain, a maximum value in a window, a variance, a standard deviation, a mode, a zero-crossing frequency, a mean square value, an over-average rate, and the like. The frequency domain analysis is to convert the time domain signal into a frequency domain signal by fourier transform, and the common frequency domain features include a peak value, a frequency component, a dc component, energy, a shape mean, a shape standard deviation, a shape skewness, a shape kurtosis, an amplitude mean, an amplitude standard deviation, an amplitude skewness, an amplitude kurtosis, a median frequency, a mean frequency, a frequency range, a highest peak frequency, a highest peak amplitude, and the like. The time domain and frequency domain features may also be collectively referred to as signal identification features.
(2) Obtaining features related to completion of a specified action based on surface electromyographic signals
The characteristics (which may be simply referred to as action characteristics) related to the completion of the designated action include a maximum myoelectric signal at the time of completion of the designated action, a time taken to complete the designated action, and the like. With the progress of the disease of PD patients, the motor symptoms of Parkinson disease are suspended and retarded, which leads to the reduction of the strength of the electromyographic signals. In addition, when the motion symptoms of PD are relatively severe, the time taken to complete the action may be longer. Thus, features associated with the completion of a given action can also effectively distinguish between different degrees of PD patients' hand motion capabilities.
(3) And selecting effective characteristics from the plurality of signal identification characteristics and action characteristics based on the information gain to train the classifier.
First, information gains of the respective features are calculated. The information gain indicates the degree of uncertainty in the information of class Y that the information of feature X is known. The information gain of the feature a to the training data set D is denoted as g (D, a), which is the difference between the empirical entropy or information entropy H (D) of the training data set D and the empirical conditional entropy H (D | a) of the training data set D given the condition of the feature a, i.e.:
g(D,A)=H(D)-H(D|A)
the empirical entropy H (D) represents the uncertainty of classifying the data set D, and the empirical entropy H (D | a) represents the uncertainty of classifying the data set D under the condition given by the feature a, and the difference between the two is the information gain, which represents the degree of uncertainty reduction of the classification of the data set D due to the feature a. The empirical entropy h (d) can be calculated, for example, by the following formula:
where n denotes the total classification of the training data set D into n classes, piThe probability of classifying the data into the ith class is represented, and the larger the entropy H known by the above formula is, the larger the uncertainty of the classification is. The empirical conditional entropy H (D | a) can be calculated, for example, by the following formula:
where n denotes the total classification of the training data set D into n classes, pi(D | A) represents the probability of classifying the data into the ith class given the feature A, and H (D | A) characterizes the degree of uncertainty in classifying the data set D given the feature A.
After the information gain of each feature of the PD movement symptom can be calculated according to the formula, the information gains of the features are compared, the features with larger information gains are selected as the features for training the classifier, and the features are extracted from the surface electromyogram signals in the training data set to serve as sample data for training the classifier.
To further illustrate the effects of the method according to the above-described embodiment, the inventors also conducted the following tests.
First, a classifier is trained using only time-domain and frequency-domain features extracted from a surface electromyogram signal. The training data for the "fist making" and "finger making" actions are processed, for example using a support Vector machine svm (support Vector machine), a sequence minimum optimization, smo (sequential minimization) model, a decision tree model such as J48, and a random forest rf (random forest) model, respectively, the experimental results being shown in fig. 2(a) and 2 (b). For the action of 'fist making', the recognition accuracy by using a J48 classifier is highest and can reach 86.2%; for the 'finger-to-finger' action, the recognition precision by using the SMO classifier is the highest and can reach 63.7%. The results of the precision (precision), Recall (Recall) and F-score for the "finger-to-finger" action using the SMO classifier to identify the class to which it belongs (i.e., classes 0-4) are shown in fig. 3(a), with these three indices being arranged in the figure in order from left to right for each class. The results of the accuracy, recall, and F-score for the "fist making" action, using the J48 classifier to identify the class to which it belongs (i.e., classes 0-4) are shown in FIG. 3 (b). By observing the strength of the electromyographic signals, the electromyographic signals of the fingers are found to be weaker than those of the fingers which make a fist, because the muscle activation degree of the fingers is weaker than that of the fingers which make a fist, and the fingers are difficult to classify when the electromyographic signals are relatively weak, so that the differentiation of the electromyographic signals of the fingers is not obvious when the fingers make a fist, and the precision of the fingers in identification is not as high as that of the fingers which make a fist. When the classification levels are 0 and 1, the error rate is high, that is, the finger-to-finger level of a normal person and the finger-to-finger level of a patient with slightly slow movement are not easily distinguished, because the normal person also has a symptom of slightly slow movement after a large amount of work and labor. When the classification levels are 2 and 3, the accuracy is low, as shown in fig. 3 (a). This is because when the patient is performing the finger-to-finger test, the muscle activation degree of the patients with grades 2 and 3 is weak, which causes the intensity of the electromyographic signal to be small, and when the electromyographic signal with small intensity slightly changes, the result of confusion of classification is easily caused. These factors are the main reasons for the high error rate of finger classification.
Next, for the "fist making" motion, a J48 classifier is used, and for the "finger pointing" motion, an SMO classifier is used, but the classifier is trained using only motion features such as the maximum electromyographic signal at the time of completion of the specified motion, and the time taken to complete the specified motion. When the J48 classifier is used, the classification accuracy of the fist classifier reaches 70.6%, and when the SMO classifier is used, the identification accuracy of the finger classifier reaches 72.5%.
And finally, selecting the first N effective characteristics from the signal identification characteristics and the action characteristics by the information gain calculation mode to train a classifier, and adopting the same classifier model as the above. The precision of the fist classifier reaches 90.8% as when the J48 classifier is used. When the SMO classifier is used, the precision of the finger-to-finger classifier reaches 82.3%. Such a designated action classifier has recognition accuracy, recall rate and F-score for the hand capability levels 0-4 as shown in FIGS. 4(a) - (b), where for each level, these three indices are arranged in the figure in order from left to right. Fig. 5(a) - (b) are graphs comparing experimental results of identifying features, action features and fusion features with signals, respectively. As shown in fig. 5, compared to using only the signal recognition feature or only the motion feature, the classification accuracy after selecting the effective feature (referred to as the fusion feature in the figure) by using the information gain is significantly improved.
Although the present invention has been described by way of preferred embodiments, the present invention is not limited to the embodiments described herein, and various changes and modifications may be made without departing from the scope of the present invention.
Claims (5)
1. A system for quantitative assessment of parkinsonian hand motor abilities, the system comprising:
the device comprises a collecting device, a hand-operated sensor and a hand-operated control device, wherein the collecting device is used for collecting a surface electromyographic signal of a detected person when the detected person performs a specified action through an electromyographic sensor placed on the hand of the detected person;
the detection device is used for judging the grade of the hand movement ability of the detected person according to the collected surface electromyographic signals by utilizing a pre-trained Parkinson patient movement ability classifier corresponding to the specified action;
wherein the features for training the parkinsonian movement ability classifier corresponding to a specified action comprise extracting time domain features and frequency domain features from a surface electromyographic signal and features related to the completion of the specified action obtained based on the surface electromyographic signal, wherein the features related to the completion of the specified action comprise at least one of: the maximum electromyographic signal when the designated action is finished and the time for finishing the designated action;
wherein the system further comprises: training means for training a parkinsonian motor ability classifier corresponding to a specified action, configured to:
a) a plurality of surface electromyographic signals which are acquired by electromyographic sensors worn by the Parkinson patients and normal persons in different degrees when the Parkinson patients and the normal persons perform the specified actions are taken as training data sets;
b) extracting time domain characteristics and frequency domain characteristics from each surface electromyographic signal and acquiring characteristics related to the completion condition of the specified action based on the surface electromyographic signal;
c) selecting features for training the athletic performance classifier based on an information gain of each feature to a training data set, wherein the information gain of each feature to the training data set is a difference between an empirical entropy of the training data set and an empirical conditional entropy of the training data set given the feature;
d) training the athletic performance classifier based on the selected features.
2. The system of claim 1, wherein the specified action comprises at least one of: fist making and finger pointing.
3. The system of claim 2, wherein when the specified action is a finger-pair, a sequence-minimum optimization model is employed as the athletic performance classifier; when the designated action is a fist, a J48 classifier is adopted as the athletic ability classifier.
4. The system of claim 1, wherein the empirical entropy of the training data set represents an uncertainty in classifying the training data set, calculated as follows:
where D represents the training data set, H (D) represents the empirical entropy of the training data set D, n represents the total classification of the training data set D into n classes, piRepresenting the probability of classifying the data into class i.
5. The system of claim 4, wherein the empirical conditional entropy of the training data set given a feature represents the uncertainty of classifying the training data set given the feature, calculated as follows:
where A denotes a certain feature, H (D | A) is the empirical entropy of the training data set D given the feature A, pi(D | A) represents the probability of classifying the data into class i given the feature A.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710463520.XA CN107157450B (en) | 2017-06-19 | 2017-06-19 | Quantitative assessment method and system for hand motion ability of Parkinson's disease people |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710463520.XA CN107157450B (en) | 2017-06-19 | 2017-06-19 | Quantitative assessment method and system for hand motion ability of Parkinson's disease people |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107157450A CN107157450A (en) | 2017-09-15 |
CN107157450B true CN107157450B (en) | 2020-03-31 |
Family
ID=59818815
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710463520.XA Active CN107157450B (en) | 2017-06-19 | 2017-06-19 | Quantitative assessment method and system for hand motion ability of Parkinson's disease people |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107157450B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108309295A (en) * | 2018-02-11 | 2018-07-24 | 宁波工程学院 | A kind of arm muscular strength assessment method |
CN108670713A (en) * | 2018-04-23 | 2018-10-19 | 中国科学院深圳先进技术研究院 | A kind of exoskeleton robot |
CN108742612B (en) * | 2018-05-24 | 2023-12-12 | 王开亮 | DBS effectiveness detection equipment based on myoelectric marker |
CN109009148A (en) * | 2018-08-24 | 2018-12-18 | 广东工业大学 | A kind of gait function appraisal procedure |
CN110547807A (en) * | 2019-09-17 | 2019-12-10 | 深圳市赛梅斯凯科技有限公司 | driving behavior analysis method, device, equipment and computer readable storage medium |
CN110516762B (en) * | 2019-10-10 | 2022-11-15 | 深圳大学 | Muscle state quantitative evaluation method and device, storage medium and intelligent terminal |
CN111210912A (en) * | 2020-01-14 | 2020-05-29 | 上海恩睦信息科技有限公司 | Parkinson prediction method and device |
CN111415746A (en) * | 2020-04-22 | 2020-07-14 | 上海邦邦机器人有限公司 | Physical function evaluation model generation method, physical function evaluation method, and physical function evaluation apparatus |
CN111528842B (en) * | 2020-05-26 | 2023-01-03 | 复嶂环洲生物科技(上海)有限公司 | Quantitative assessment method for Parkinson disease symptoms based on physiological and behavioral indexes |
CN112826504B (en) * | 2021-01-07 | 2024-03-26 | 中新国际联合研究院 | Game parkinsonism grade assessment method and device |
CN117643456B (en) * | 2024-01-29 | 2024-05-24 | 北京航空航天大学 | Auxiliary evaluation system, method and storage medium for parkinsonism |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050234309A1 (en) * | 2004-01-07 | 2005-10-20 | David Klapper | Method and apparatus for classification of movement states in Parkinson's disease |
US9060714B2 (en) * | 2008-12-04 | 2015-06-23 | The Regents Of The University Of California | System for detection of body motion |
CN104434129B (en) * | 2014-12-25 | 2016-08-17 | 中国科学院合肥物质科学研究院 | It is that disease movement disorder symptoms quantifies evaluating apparatus and method outside a kind of parkinson and relevant cone |
CN104398263B (en) * | 2014-12-25 | 2018-02-16 | 中国科学院合肥物质科学研究院 | A kind of disturbances in patients with Parkinson disease based on approximate entropy and mutual approximate entropy tremble symptom quantify evaluating method |
CN104522949B (en) * | 2015-01-15 | 2016-01-06 | 中国科学院苏州生物医学工程技术研究所 | A kind of Intelligent bracelet for qualitative assessment disturbances in patients with Parkinson disease motor function |
CN105426842B (en) * | 2015-11-19 | 2018-08-14 | 浙江大学 | Multiclass hand motion recognition method based on support vector machines and surface electromyogram signal |
CN105550583B (en) * | 2015-12-22 | 2018-02-13 | 电子科技大学 | Android platform malicious application detection method based on random forest classification method |
CN105426696A (en) * | 2015-12-24 | 2016-03-23 | 中国科学院苏州生物医学工程技术研究所 | Multi-node quantitative assessment system and method for symptoms of Parkinson's disease |
CN105930663B (en) * | 2016-04-26 | 2020-06-19 | 北京科技大学 | Hand tremor signal and audio signal classification method based on evolution fuzzy rule |
-
2017
- 2017-06-19 CN CN201710463520.XA patent/CN107157450B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN107157450A (en) | 2017-09-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107157450B (en) | Quantitative assessment method and system for hand motion ability of Parkinson's disease people | |
Burns et al. | Upper limb movement classification via electromyographic signals and an enhanced probabilistic network | |
Printy et al. | Smartphone application for classification of motor impairment severity in Parkinson's disease | |
Miften et al. | A new framework for classification of multi-category hand grasps using EMG signals | |
Krishnan et al. | Recognition of human arm gestures using Myo armband for the game of hand cricket | |
Bazgir et al. | A neural network system for diagnosis and assessment of tremor in Parkinson disease patients | |
Oung et al. | Wearable multimodal sensors for evaluation of patients with Parkinson disease | |
US20200375500A1 (en) | Systems and methods for detection and correction of abnormal movements | |
CN108305680A (en) | Intelligent parkinsonism aided diagnosis method based on multi-element biologic feature and device | |
Naik et al. | Classification of finger extension and flexion of EMG and Cyberglove data with modified ICA weight matrix | |
KR20200118525A (en) | Evaluation of Parkinson's disease index using acceleration and angular velocity signals and method for evaluation thereof | |
CN111401435A (en) | Human body motion mode identification method based on motion bracelet | |
Said et al. | Machine-learning based wearable multi-channel sEMG biometrics modality for user's identification | |
Sharma et al. | On the use of temporal and spectral central moments of forearm surface EMG for finger gesture classification | |
Veer | Experimental study and characterization of SEMG signals for upper limbs | |
Louis et al. | One dimensional multi-resolution local binary patterns features (1DMRLBP) for regular electrocardiogram (ECG) waveform detection | |
CN113609975A (en) | Modeling method for tremor detection, hand tremor detection device and method | |
Tomczyński et al. | Hand gesture-based interface with multichannel sEMG band enabling unknown gesture discrimination | |
CN107198508B (en) | Recovery degree sequencing method and combined interactive training system | |
CN111297366B (en) | Data processing method and diagnosis device for assisting disease diagnosis based on daily necessities | |
Sidek et al. | A comparative analysis of QRS and cardioid graph based ECG biometric recognition in different physiological conditions | |
Yu et al. | Classification of nine directions using the maximum likelihood estimation based on electromyogram of both forearms | |
Anam et al. | Index Finger Motion Recognition using self-advise support vector machine | |
Dev et al. | Performance analysis of classifiers for EMG signal of different hand movements | |
Pathoumvanh et al. | Implementation of the ECG biometric identification by using Arduino Microprocessor |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |