TRCA-cpp-src

• class Cheby1Filter: Design of bandpass, bandstop filters

• class Preprocess: notch, FilterBank, detrend, average, std calculation

• class TRCA: TRCA implementation

• utils: data, debug function

• for Chinese, vstudio默认gb2312编码, vs code默认utf8编码, 大部分含中文注释文件已经转换为utf8编码

TRCA.dll

• For source code, refer to dll.cpp.

• For python call, refer to dllValid.py, and onlineValid.py

FilterBank

• Implementation of notch, FilterBank (10 banks, 6-th order bandpass at most: [{6, 14, 22, 30, 38, 46, 54, 62, 70, 78}, 90]), detrend, normalization.

• We use Cheby1 to design all filters.

• Input(2 pointers, 7 int)

  • darray: double* (row major 4D array, [block, stimulus, electrode, samples]). Generally, block=1, stimulus=the number of trials.

  • dout: double* (row major 4D array, [block*stimulus, subbands number, electrode, samples])

  • s_rate: int (sampling rate)

  • subbands: int (the number of subbands in FilterBank)

  • len: int (the number of blocks)

  • stimulus: int (the number of stimulus)

  • electrodes: int (the number of electrodes)

  • num_samples: int (the number of samples)

  • debug: int (=1 to debug mode, the data of FilterBank will be written into ./FilterBank.csv)

• Output

  • dout: memcpy to dout*

  • exit code (TODO)

TrcaTrain

TRCA training (=FilterBank + TrcaTrainOnly)

• Input (3 pointers, 7 int)

  • darray: double* (row major 4D array, [train blocks, stimulus, electrodes, samples])

  • pTemplate: double* (4D array, [stimulus, subbans, electrodes, samples]). Generally, we don't consider it is row major or col major, since it will be directly used in TrcaTest. This function will output col major data, The same applies hereinafter.

  • pU: double* (4D array, [subbands, stimulus, electrodes, 1]). Generally, we don't consider it is row major or col major, since it will be directly used in TrcaTest. This function will output col major data, The same applies hereinafter.

  • s_rate: int (sampling rate)

  • subbands: int (the number of subbands in FilterBank)

  • train_len: int (the number of train blocks)

  • stimulus: int (the number of stimulus)

  • electrodes: int (the number of electrodes)

  • num_samples: int (the number of samples)

  • debug: int (=1 to csv_output mode, templates, and u will be written into ./*.csv. =2 to debug mode, the data of FilterBank, templates, and u will be written into ./*.csv.)

• Output

  • pTemplate: memcpy to pTemplate*

  • pU: memcpy to pU

  • exit code (TODO)

TrcaTrainOnly

Only train of TRCA is implemented, without FilterBank.

• Input (3 pointers, 7int)

  • darray: double* (row major 4D array, [train blocks*stimulus, subbands, electrodes, samples]), which is the output of FilterBank

  • pTemplate: double* (4D array, [stimulus, subbands, electrodes, samples])

  • pU: double* (4D array, [subbands, stimulus, electrodes, 1])

  • s_rate: int (sampling rate)

  • subbands: int (the number of subbands)

  • train_len: int (the number of train blocks)

  • stimulus: int (the number of stimulus)

  • electrodes: int (the number of electrodes)

  • num_samples: int (the number of samples)

  • debug: int (=1 to csv_output mode, templates, and u will be written to ./*.csv. =2 to debug mode, input, templates, and u will be written to ./*.csv)

• Output

  • pTemplate: memcpy to pTemplate*

  • pU: memcpy to pU*

  • exit code (TODO)

TrcaTest

TRCA testing (=FilterBank + TrcaTestOnly)

• Input (5 pointers, 7 int)

  • darray: double* (row major 4D array, [1, test blocks, electrodes, samples])

  • pTemplate: double* (output of TrcaTrain)

  • pU: double*, (output of TrcaTrain)

  • pcoeff: double* (1D array, [test blocks*stimulus])

  • pPred: int* (1D array, [test blocks])

  • s_rate: int (sampling rate)

  • subbands: int (the number of subbands)

  • test_len: int (the number of test blocks)

  • stimulus: int (the number of stimulus)

  • electrodes: int (the number of electrodes)

  • num_samples: int (the number of samples)

  • debug: int (=1 to debug mode, templates, and u will be written to ./*.csv)

• Output

  • pPred: memcpy to

  • pcoeff: memcpy to

  • exit code (TODO)

TrcaTestOnly

Only test of TRCA is implemented, without FilterBank.

• Input (5 pointers, 7 int)

  • darray: double* (row major 4D array, [1*test blocks, subbands, electrodes, samples])

  • pTemplate: double*, (output of TrcaTrain)

  • pU: double*, (output of TrcaTrain)

  • pcoeff: double* (1D array, [test blocks*stimulus])

  • pPred: int* (1D array, [test blocks])

  • s_rate: int (sampling rate)

  • subbands: int (the number of subbands)

  • test_len: int (the number of test blocks)

  • stimulus: int (the number of stimulus)

  • electrodes: int (the number of electrodes)

  • num_samples: int (the number of samples)

  • debug: int (=1 to debug mode, templates, u ,and input will be written to ./*.csv)

• Output

  • pPred: memcpy to pPred*

  • pcoeff: memcpy to pcoeff*

  • exit code (TODO)

TrcaTestCsv

TRCA testing (=FilterBank + TrcaTestOnly), which use template.csv and u.csv for input

• Input (5 pointers, 7 int)

  • darray: double* (row major 4D array, [1, test blocks, electrodes, samples])

  • pTemplate: char*, path to templates.csv

  • pU: char*, path to u.csv

  • pcoeff: double* (1D array, [test blocks*stimulus])

  • pPred: int* (1D array, [test blocks])

  • s_rate: int (sampling rate)

  • subbands: int (the number of subbands)

  • test_len: int (the number of test blocks)

  • stimulus: int (the number of stimulus)

  • electrodes: int (the number of electrodes)

  • num_samples: int (the number of samples)

  • debug: int (=1 to debug mode, templates, and u will be written to ./*.csv)

• Output

  • pPred: memcpy to pPred*

  • pcoeff: memcpy to pcoeff*

  • exit code (TODO)

TrcaTestOnlyCsv

TRCA testing, without FilterBank, which use template.csv and u.csv for input

• Input (5 pointers, 7 int)

  • darray: double* (row major 4D array, [1, test blocks, electrodes, samples])

  • pTemplate: char*, path to templates.csv

  • pU: char*, path to u.csv

  • pcoeff: double* (1D array, [test blocks*stimulus])

  • pPred: int* (1D array, [test blocks])

  • s_rate: int (sampling rate)

  • subbands: int (the number of subbands)

  • test_len: int (the number of test blocks)

  • stimulus: int (the number of stimulus)

  • electrodes: int (the number of electrodes)

  • num_samples: int (the number of samples)

  • debug: int (=1 to debug mode, templates, and u will be written to ./*.csv)

• Output

  • pPred: memcpy to pPred*

  • pcoeff: memcpy to pcoeff*

  • exit code (TODO)

Application

• dllvalid.py: use [SSVEP-AnaTool] (https://github.com/pikipity/SSVEP-Analysis-Toolbox) for testing. Need to check the version of toolbox, or use toolbox in this repo.

• Use Wearable-SSVEP (wet) dataset for testing. When using SSVEPAnalysisToolbox, it is better to check if get_data function in the toolbox can get correct data segments. Need to compare the toolbox code and the dataset documentation.

• If data rearrange is needed, we suggest use loop to manually do that, rather than transpose or reshpae in third-party lib.