def extract(self, path):
index_of_sirens = []
index_of_not_sirens = []
rows = []
files_frequencies_ar = []
time = []
sm = specgram_maker.SpecgramMaker()
directory = os.listdir(path)
l = 0 # used to count the index of the files.
for filename in directory:
if not filename.endswith(".wav"):
continue
if filename.__contains__("siren"):
index_of_sirens.append(l)
else:
index_of_not_sirens.append(l)
spec, freq, t = sm.get_specgram_data_from_wav(path, filename)
for i in range(len(spec[1])): # iterating over coloums.
max_frequency_value = 0
row = 0
for j in range(FREQUENCY_INTEVALS
): # Finding the row with highest frequency.
if spec[j][i] > max_frequency_value:
max_frequency_value = spec[j][i]
row = j
rows.append(50 * row)
files_frequencies_ar.append(rows.copy())
rows.clear()
time.append(t)
l += 1
return files_frequencies_ar, time, index_of_sirens, index_of_not_sirens
def _find_vertical_maxima(path,
directory,
min_freq=750,
max_freq=1600,
split=True,
training=True):
specmaker = sm.SpecgramMaker()
labels = []
directory = os.listdir(path)
filenames = []
time = []
rows = []
files_frequencies_array = []
for filename in directory:
# We only consider .wav-files
if not filename.endswith(".wav"):
continue
filenames.append(filename)
# We append whether or not the filename says there is a siren in the clip. There have been many ways to
# indicate this, so this line is kind of long. :(
labels.append("sirenAt" in filename or "siren" in filename
or "SPCSiren" in filename)
# We convert the .wav-file into a matrix with specmaker.
spec, freq, t = specmaker.get_specgram_data_from_wav(path, filename)
# For each column:
for col in range(len(spec[1])):
max_dB = 0
max_row = 0
# For each row:
for row in range(int(min_freq / 50), int(
max_freq /
50)): # We divide with 50, because the frequencies
# are in multiples of 50.
# We find the highest value in the column.
if spec[row][col] > max_dB:
max_dB = spec[row][col]
max_row = row
# When we have found the highest column, we append that number times 50.
# This is because the y-axis works in multiples of 50.
rows.append([50 * max_row, max_dB])
# files_frequencies_array now contains rows, which is a list of the highest values in each column.
# I'm not sure, why the rows are copied and then cleared.
files_frequencies_array.append(rows.copy())
rows.clear()
# We would like to keep track of the times for future use.
time.append(t)
# Here we print, to keep track of how long we are in importing.
print("Done importing " + filename + ".", labels[len(time) - 1],
str(len(time) * 100 / len(directory)) + " %")
# This is where we split,, if split is true
if split:
# We call different methods dependant on whether it is used for training or not.
if training:
split_waves, labels = _split_training(files_frequencies_array,
labels, filenames, time,
divisions)
else:
split_waves, labels = _split_testing(files_frequencies_array,
labels, filenames, time,
divisions)
# If we don't split, we simply copy it into split_waves.
else:
split_waves = files_frequencies_array
new_waves = []
for feature_set in split_waves:
new_feature_set = []
for tubel_element in feature_set:
new_feature_set.append(tubel_element[0])
new_feature_set.append(tubel_element[1])
new_waves.append(new_feature_set)
return new_waves, time, labels
import clip_split
import matplotlib.pyplot as plot
import specgram_maker as sm
import os
if __name__ == "__main__":
spec = sm.SpecgramMaker()
directory = "C:\\Users\\Magnus\\Desktop\\GetPatterns\\"
for filename in os.listdir(directory):
matrix, freq, t = spec.get_specgram_data_from_wav(directory, filename)
min_freq = 14
max_freq = 32
rows = []
for col in range(len(matrix[1])): # iterating over coloums.
max_dB = 0
max_row = 0
for row in range(
min_freq,
max_freq): # Finding the row with highest frequency.
if matrix[row][col] > max_dB:
max_dB = matrix[row][col]
max_row = row
rows.append(50 * max_row)
plot.scatter(t, rows)
plot.xlabel("time / seconds")
plot.ylabel("Frequency / Hz")
plot.title(filename)
plot.show()
if __name__ == "__main__":
f = open("Log.txt", "wb")
# GPIO.setwarnings(False) # Ignore warning for now
# GPIO.setmode(GPIO.BOARD) # Use physical pin numbering
# GPIO.setup(8, GPIO.OUT, initial=GPIO.LOW) # Set the light to off as default
is_recording = True # A boolean value that runs the while loop below.
# gau = pickle.load(open("Jamesnb.v.1.pkl", "rb"))
lr = pickle.load(open("Jameslr.v.1.pkl", "rb"))
ran = pickle.load(open("JamesrandomForest.v.1.pkl", "rb"))
# svm = pickle.load(open("Jamessvm.v.1.pkl", "rb"))
tree = pickle.load(open("Jamestree.v.1.pkl", "rb"))
models = [[lr, "logistic regression"], [tree, "decision tree"],
[ran, "random forest"]]
mic = MicRecorder(20 / 6)
stream = mic.get_stream()
sgm = specgram_maker.SpecgramMaker()
while is_recording: # Vi bør finde en måde at gøre den False på
rows = []
predictions = []
print("Begin recording")
matrix, freq, t = sgm.make_specgram_from_mic_matrix(mic, stream)
print("End recording")
for col in range(len(matrix[1])): # iterating over coloums.
max_dB = 0
max_row = 0
for row in range(14,
32): # Finding the row with highest frequency.
if matrix[row][col] > max_dB:
max_dB = matrix[row][col]
max_row = row