[Speed Processing] Processing Praat TextGrids to calculate F1, F2 and F0 of angry and sad emotion files.

praat_speech_processing.py

import glob
import tgt
import sys
import parselmouth
import statistics

DEBUG = False

def get_f0_mean_stdev(filesSounds):
    sum = 0
    f0s = []
    for i, filename in enumerate(filesSounds):
        pitch = parselmouth.Sound(filename).to_pitch()
        f0_val = pitch.get_mean_absolute_slope()
        sum += f0_val
        f0s.append(f0_val)

    return [sum/len(filesSounds), statistics.stdev(f0s)]

def process_formants(filesTextgrid, filesSound):
    vowels_formants = {
        "a": [[],[]],
        "i": [[],[]],
        "u": [[],[]],
    }

    for i, filename in enumerate(filesTextgrid):
        try:
            tg = tgt.io.read_textgrid(filename)
            formant = parselmouth.Sound(filesSound[i]).to_formant_burg()

            if DEBUG:
                print(filename + " + " + filesSound[i])

            for tiername in tg.get_tier_names():
                if tiername in ["a", "i", "u"]:
                    annotation = tg.get_tier_by_name(tiername).get_annotations_with_text(tiername)[0]
                    
                    f1_start_value = formant.get_value_at_time(1, annotation.start_time)
                    f1_end_value = formant.get_value_at_time(1, annotation.end_time)
                    f1_mean = (f1_start_value + f1_end_value) / 2

                    f2_start_value = formant.get_value_at_time(2, annotation.start_time)
                    f2_end_value = formant.get_value_at_time(2, annotation.end_time)
                    f2_mean = (f2_start_value + f2_end_value) / 2

                    vowels_formants[tiername][0].append(f1_mean)
                    vowels_formants[tiername][1].append(f2_mean)
                    
                    if DEBUG:
                        print(filename + " " + format_formant(tiername, f1_mean, f2_mean))
                elif DEBUG:
                    print('\t Skipping: ' + tiername)
        except err:
            print(filename + ' caused a problem.')
            sys.stderr.write('ERROR: %s\n' % str(err))
    return vowels_formants

def calculateFormantAverages(v_formants):
    vowels = {
        "a": [0,0],
        "i": [0,0],
        "u": [0,0],
    }
 
    for vowel_name, formants in v_formants.items():
        sum = 0
        for f0 in formants[0]:
            sum += f0
            vowels[vowel_name][0] = sum / len(formants[0])

        sum = 0
        for f1 in formants[1]:
            sum += f1
            vowels[vowel_name][1] = sum / len(formants[1])

    return vowels

def print_formants(formants):
    for vowel_name, formants in formants.items():
        print(format_formant(vowel_name, formants[0], formants[1]))

def format_formant(vowel, f1, f2):
    return "\t" + vowel + ") F1: " + str(f1) + " | F2: " + str(f2)

def main(argv=None):
    print("=== Angry Emotions Analysis ===")
    print("===============================\n")
    filesTextgrid = glob.glob(r'../A_*.TextGrid')
    filesSound = glob.glob(r'../wav/A_*.wav')
    angry_formants = process_formants(filesTextgrid, filesSound)
    print_formants(calculateFormantAverages(angry_formants))
    mean, stdev = get_f0_mean_stdev(filesSound)
    print("\tF0) Mean: " + str(mean) + " | Stdev: " + str(stdev))

    print("\n=== Sad Emotions Analysis ===")
    print("=============================\n")
    filesTextgrid = glob.glob(r'../S_*.TextGrid')
    filesSound = glob.glob(r'../wav/S_*.wav')
    sad_formants = process_formants(filesTextgrid, filesSound)
    print_formants(calculateFormantAverages(sad_formants))
    mean, stdev = get_f0_mean_stdev(filesSound)
    print("\tF0) Mean: " + str(mean) + " | Stdev: " + str(stdev))

if __name__ == '__main__':
    sys.exit(main())