open source pkg v1

dbm_lib/dbm_features/raw_features/audio/gne.py

@@ -0,0 +1,157 @@
+"""
+file_name: gne
+project_name: DBM
+created: 2020-20-07
+"""
+
+import pandas as pd
+import numpy as np
+import os
+import glob
+import parselmouth
+import librosa
+import more_itertools as mit
+from os.path import join
+import logging
+
+from dbm_lib.dbm_features.raw_features.util import util as ut
+
+logging.basicConfig(level=logging.INFO)
+logger=logging.getLogger()
+
+gne_dir = 'audio/glottal_noise'
+ff_dir =  'audio/pitch'
+csv_ext = '_gne_frame.csv'
+
+def gne_ratio(sound):
+    """
+    Using parselmouth library fetching glottal noise excitation ratio
+    Args:
+        sound: parselmouth object
+    Returns:
+        (list) list of gne ratio for each voice frame
+    """
+    harmonicity_gne = sound.to_harmonicity_gne()
+    gne_all_bands = harmonicity_gne.values
+    gne_all_bands = np.where(gne_all_bands==-200, np.NaN, gne_all_bands)
+    
+    gne = np.nanmax(gne_all_bands) # following http://www.fon.hum.uva.nl/rob/NKI_TEVA/TEVA/HTML/NKI_TEVA.pdf
+    return gne
+
+def empty_gne(video_uri, out_loc, fl_name, r_config, error_txt):
+    """
+    Preparing empty GNE matrix if something fails
+    """
+    cols = ['Frames', r_config.aco_gne, r_config.err_reason]
+    out_val = [[np.nan, np.nan, error_txt]]
+    
+    df_gne = pd.DataFrame(out_val, columns = cols)
+    df_gne['dbm_master_url'] = video_uri
+    
+    logger.info('Saving Output file {} '.format(out_loc))
+    ut.save_output(df_gne, out_loc, fl_name, gne_dir, csv_ext)
+    
+def segment_pitch(dir_path, r_config):
+    """
+    segmenting pitch freq for each voice segment
+    """
+    com_speech_sort, voiced_yes, voiced_no  = ([], ) * 3
+    for file in os.listdir(dir_path):
+        try:
+            
+            if file.endswith('_pitch.csv'):
+                
+                ff_df = pd.read_csv((dir_path+'/'+file))
+                voice_label = ff_df[r_config.aco_voiceLabel]
+                
+                indices_yes = [i for i, x in enumerate(voice_label) if x == "yes"]
+                voiced_yes = [list(group) for group in mit.consecutive_groups(indices_yes)]
+                
+                indices_no = [i for i, x in enumerate(voice_label) if x == "no"]
+                voiced_no = [list(group) for group in mit.consecutive_groups(indices_no)]
+                
+                com_speech = voiced_yes + voiced_no
+                com_speech_sort = sorted(com_speech, key=lambda x: x[0])
+        except:
+            pass
+        
+    return com_speech_sort, voiced_yes, voiced_no
+
+def segment_gne(com_speech_sort, voiced_yes, voiced_no, gne_all_frames, audio_file):
+    """
+    calculating gne for each voice segment
+    """
+    snd = parselmouth.Sound(audio_file)
+    pitch = snd.to_pitch(time_step=.001)
+    
+    for idx, vs in enumerate(com_speech_sort):
+        try:
+            
+            max_gne = np.NaN
+            if vs in voiced_yes and len(vs)>1:
+                
+                start_time = pitch.get_time_from_frame_number(vs[0])
+                end_time = pitch.get_time_from_frame_number(vs[-1])
+
+                snd_start = int(snd.get_frame_number_from_time(start_time))
+                snd_end = int(snd.get_frame_number_from_time(end_time))
+
+                samples = parselmouth.Sound(snd.as_array()[0][snd_start:snd_end])
+                max_gne = gne_ratio(samples)
+        except:
+            pass
+
+        gne_all_frames[idx] = max_gne
+    return gne_all_frames
+    
+def calc_gne(video_uri, audio_file, out_loc, fl_name, r_config):
+    """
+    Preparing gne matrix
+    Args:
+        audio_file: (.wav) parsed audio file
+        out_loc: (str) Output directory for csv's
+    """
+    dir_path = os.path.join(out_loc, ff_dir)
+    if os.path.isdir(dir_path):
+        voice_seg = segment_pitch(dir_path, r_config)
+        
+        gne_all_frames = [np.NaN] * len(voice_seg[0])
+        gne_segment_frames = segment_gne(voice_seg[0], voice_seg[1], voice_seg[2], gne_all_frames, audio_file)
+        
+        df_gne = pd.DataFrame(gne_segment_frames, columns=[r_config.aco_gne])
+        df_gne[r_config.err_reason] = 'Pass'# will replace with threshold in future release
+        
+        df_gne['Frames'] = df_gne.index
+        df_gne['dbm_master_url'] = video_uri
+        
+        logger.info('Processing Output file {} '.format(out_loc))
+        ut.save_output(df_gne, out_loc, fl_name, gne_dir, csv_ext)
+        
+    else:
+        error_txt = 'error: pitch freq not available'
+        empty_gne(video_uri, out_loc, fl_name, r_config, error_txt)
+
+def run_gne(video_uri, out_dir, r_config):
+    """
+    Processing all patient's for fetching glottal noise ratio
+    ---------------
+    ---------------
+    Args:
+        video_uri: video path; r_config: raw variable config object
+        out_dir: (str) Output directory for processed output
+    """
+    input_loc, out_loc, fl_name = ut.filter_path(video_uri, out_dir)
+    aud_filter = glob.glob(join(input_loc, fl_name + '.wav'))
+    if len(aud_filter)>0:
+        
+        audio_file = aud_filter[0]
+        aud_dur = librosa.get_duration(filename=audio_file)
+
+        if float(aud_dur) < 0.064:
+            logger.info('Output file {} size is less than 0.064sec'.format(audio_file))
+            
+            error_txt = 'error: length less than 0.064'
+            empty_gne(video_uri, out_loc, fl_name, r_config, error_txt)
+            return
+
+        calc_gne(video_uri, audio_file, out_loc, fl_name, r_config)