characterDefinitions.py

#--Can add new definitions here to apply this code to other problems or datasets.--
import numpy as np

def getHandwritingCharacterDefinitions():
    """
    Returns a dictionary with entries that define the names of each character, its length, and where the pen tip begins.
    
    Returns:
        charDef (dict)
    """
        
    charDef = {}
    
    #Define the list of all 31 characters and their names.
    charDef['charList'] = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z',
                'greaterThan','comma','apostrophe','tilde','questionMark']
    charDef['charListAbbr'] = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z',
                '>',',',"'",'~','?']

    #Define the length of each character (in # of 10 ms bins) to use for each template.
    #These were hand-defined based on visual inspection of the reconstructed pen trajectories.
    charDef['charLen'] = np.array([99, 91, 70, 104, 98, 125, 110, 104, 79, 92, 127, 68, 132, 90, 
                        84, 113, 104, 74, 86, 110, 86, 83, 110, 103, 115, 100, 82, 77, 116, 71, 110]).astype(np.int32)
    
    #For each character, this defines the starting location of the pen tip (0 = bottom of the line, 1 = top)
    charDef['penStart'] = [0.25, 1, 0.5, 0.5, 0.25, 1.0, 0.25, 1.0, 0.5, 0.5, 1, 1, 0.5, 0.5, 0.25, 0.5, 0.25, 0.5, 0.5, 1, 
           0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.25, 1, 0.5, 1]
    
    #dictionary to convert string representation to character index
    charDef['strToCharIdx'] = {}
    for x in range(len(charDef['charListAbbr'])):
        charDef['strToCharIdx'][charDef['charListAbbr'][x]] = x
        
    #ordering of characters that kaldi (i.e., the language model) expects
    charDef['kaldiOrder'] = ['<ctc>','>',"'",',','.','?','a','b','c','d','e','f','g','h','i','j',
                             'k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z']
    
    #re-indexing to match kaldi order (e.g., outputs[:,:,charDef['idxToKaldi']] places the output in kald-order)
    charDef['idxToKaldi'] = np.array([31,26,28,27,29,30,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,
                                     21,22,23,24,25]).astype(np.int32)
    
    return charDef