README.md

Unlabeled_STM

This sample scripts are described in the paper "Subject-transfer framework with unlabeled data based on multiple distance measures for surface electromyogram pattern recognition" accepeted to Biomedical Signal Processing and Control.

<Description>
After changing information about your directories in main_script.m (lines 6 and 8), downloading getxxfeat.m, and installing LIBSVM package, you can use this codes.
This project has three folders:

1. data
   

   • private
     
     • 22-class (1-DoF 8-class and 2-DoF 14-class) EMG data from 25 subjects
       
       • The detail is described in here
         
       • We only used 1-DoF 8-class data (i.e., from M1T1.csv to M8T5.csv)
         
     • csv files (each data has 5-s information: the last 1 s is already cut)
       
     • M means the motion label (e.g., M1 means resting state and M2 means wrist flexion)
       
     • T means the number of trials
       
     • After applying preprocessing_ds1.m, F_c.mat will be added in the folter .../data/private.
       
   • NinaPro DB5 exerciseA
     
     • 12-class finger motions from 10 subjects
       
       • You can get this datasets from here
         
     • Put SX_E1_A1.mat in each subject's folder
       
     • S means the subject
       
     • After applying preprocessing_ds2.m, F_c.mat will be added in the folter .../data/NinaPro DB5 exerciseA
       
   • NinaPro DB5 exerciseB
     
     • 17-class hand and wrist motions from 10 subjects
       
     • Put SX_E2_A1.mat in each subject's folder
       
     • S means the subject
       
     • After applying preprocessing_ds3.m, F_c.mat will be added in the folter .../data/NinaPro DB5 exerciseB
       
   • NinaPro DB5 exerciseC
     
     • 23-class hand and wrist motions from 10 subjects
       
     • Put SX_E3_A1.mat in each subject's folder
       
     • S means the subject
       
     • After applying preprocessing_ds4.m, F_c.mat will be added in the folter .../data/NinaPro DB5 exerciseC
       

2. code
   this folder has one main m.file named main_script that uses:
   

   • set_config
     
   • preprocessing_ds1
     
     • extract_features
       you can get the following m.files from here
       
       • getrmsfeat
         
       • getmavfeat
         
       • getzcfeat
         
       • getsscfeat
         
       • getwlfeat
         
       • getarfeat
         
   • preprocessing_ds2
     
     • extract_features
       
       • getrmsfeat
         
       • getmavfeat
         
       • getzcfeat
         
       • getsscfeat
         
       • getwlfeat
         
       • getarfeat
         
   • preprocessing_ds3
     
     • extract_features
       
       • getrmsfeat
         
       • getmavfeat
         
       • getzcfeat
         
       • getsscfeat
         
       • getwlfeat
         
       • getarfeat
         
   • preprocessing_ds4
     
     • extract_features
       
       • getrmsfeat
         
       • getmavfeat
         
       • getzcfeat
         
       • getsscfeat
         
       • getwlfeat
         
       • getarfeat
         
   • evaluate_lda_acc
     
   • evaluate_svm_acc
     you can make the m scripts, svmtrain.m and svmpredict.m from here
     
     • svmtrain (LIBSVM)
       
     • svmpredict (LIBSVM)
       
     • supervised_STM
       
       • find_target
         
       • calculate_A_b
         
   • evaluate_lda_mdms
     
   • evaluate_svm_mdms
     
     • svmtrain (LIBSVM)
       
     • svmpredict (LIBSVM)
       
     • supervised_STM
       
       • find_target
         
       • calculate_A_b
         
   • evaluate_lda_random
     
   • evaluate_svm_random
     
     • svmtrain (LIBSVM)
       
     • svmpredict (LIBSVM)
       
   • visualize_results
     

3. resuts
   this folder will store results_lda/svm_acc/mdms/random_dsx.mat and boxplots.fig.
   

<Environments>
Windows 10
MATLAB R2020a
1. Signal Processing Toolbox
2. Statics and Machine Learning Toolbox
3. Parallel Comupting Toolbox