def train_classifier(dataset,n_classes,Fs=16000):
datalist = os.listdir(dataset)
n_data_set = len(datalist)
for i in range(n_data_set) :
filepath = dataset + '/' + datalist[i]
if datalist[i].find('.wav')==1 :
try:
[x, Fs, n_channels, n_samples] = read_wave(filepath)
except:
print(e.msg)
elif datalist[i].find('.raw') :
try:
x = read_raw(filepath)
if len(x.shape)>1:
n_channels = x.shape[1]
n_samples = x.shape[0]
else:
n_channels = 1
n_samples = len(x)
except:
print(e.msg)
features = get_mfcc(x,Fs)
dimension = features.shape[1]
size = features.shape[0]
emHMM_algorithm(features,dimension,2,size)
def train_classifier(dataset, n_classes, Fs=16000):
datalist = os.listdir(dataset)
n_data_set = len(datalist)
for i in range(n_data_set):
filepath = dataset + '/' + datalist[i]
if datalist[i].find('.wav') == 1:
try:
[x, Fs, n_channels, n_samples] = read_wave(filepath)
except:
print(e.msg)
elif datalist[i].find('.raw'):
try:
x = read_raw(filepath)
if len(x.shape) > 1:
n_channels = x.shape[1]
n_samples = x.shape[0]
else:
n_channels = 1
n_samples = len(x)
except:
print(e.msg)
features = get_mfcc(x, Fs)
dimension = features.shape[1]
size = features.shape[0]
emHMM_algorithm(features, dimension, 2, size)
def dir_scan_azimuth(scan_folder, elevation, result):
scan_str = 'H' + str(elevation) + 'e'
for file in glob.glob(scan_folder + "/" + scan_str + "*.wav"):
azimuthstrpos = string.find(file, scan_str)
if azimuthstrpos > 0:
azimuth = int(file[azimuthstrpos + len(scan_str):azimuthstrpos + len(scan_str) + 3])
left_c_str, right_c_str, length = read_wave(file)
result.append((elevation, azimuth, left_c_str, right_c_str, length))
def dir_scan_azimuth(scan_folder, elevation, result):
scan_str = 'H' + str(elevation) + 'e'
for file in glob.glob(scan_folder + "/" + scan_str + "*.wav"):
azimuthstrpos = string.find(file, scan_str)
if azimuthstrpos > 0:
azimuth = int(file[azimuthstrpos + len(scan_str):azimuthstrpos +
len(scan_str) + 3])
left_c_str, right_c_str, length = read_wave(file)
result.append(
(elevation, azimuth, left_c_str, right_c_str, length))
def loadSample():
print("Enter full path to .wav file")
path = input()
signal, sr = rw.read_wave(path, normalize=True, length=1, threshold=0.001)
spec = sp.get_spectrogram(signal,
sr,
frame_length=frame_length,
frame_offset=frame_offset,
lowFreq=lowFreq,
hiFreq=hiFreq,
numFilters=numFilters,
numFrames=numFrames)
return np.array([spec])
numDataPoints = numFilters * numFrames
for dir, out in zip(directories, out_files):
print("Writing " + dir + " data to: " + out)
# Get filepaths of samples for each class
class_files = {}
for c in CLASSES:
files = os.listdir(dir + '/' + c)
class_files[c] = files
# Read, process, and write data of each sample to a csv file
data = []
for c in class_files:
print("Class being processed: " + c)
files = class_files[c]
paths = []
for file in files:
paths.append(dir + '/' + c + '/' + file)
for path in tqdm(paths):
signal, sr = rw.read_wave(path, normalize=True, length=1, threshold=0.001)
spec = sp.get_spectrogram(signal, sr, frame_length=frame_length, frame_offset=frame_offset, lowFreq=lowFreq, hiFreq=hiFreq, numFilters=numFilters, numFrames=numFrames)
dataToWrite = np.append(np.array([c]), spec.flatten())
data.append(dataToWrite)
header = ['Class']
header.extend(['D' + str(i) for i in range(numDataPoints)])
df = pd.DataFrame(columns=header, data=data)
df.to_csv(out)