def predict(data):
"""
parameters
==========
data : mfcc 数组
"""
mpaths, model_name = get_model_fpath(model_dir_name)
ubm = GMM.load(ubm_path)
model = []
for mname, cname in zip(mpaths, model_name):
model.append(GMM.load(mname))
ubm_score_all = ubm.score_all(data)
score_list = []
for m in model:
score_list.append(m.score_all(data))
max_index = score_list.index(max(score_list))
return model_name[max_index]
def get_gmm():
# from sklearn.mixture import GMM as skGMM
# from gmmset import GMM as pyGMM
# if GMM == skGMM:
# print 'using GMM from sklearn'
# return GMM(nr_mixture)
# else:
# print 'using pyGMM'
return GMM(nr_mixture=nr_mixture,
nr_iteration=500,
init_with_kmeans=0,
concurrency=8,
threshold=1e-15,
verbosity=2)
def main(datapath, ubmpath, gmmPath):
fpaths = get_training_data_fpaths(datapath)
# print "The fpath is :",fpaths
X_train, y_train = datautil.read_data(fpaths)
ubm = GMM.load(ubmpath)
for x, y in zip(X_train, y_train):
gmm = GMM(concurrency=8, threshold=0.01, nr_iteration=100, verbosity=1)
start = time.time()
gmm.fit(x, ubm=ubm)
# score = gmm.score(X_train[0])
# print(gmm.weights_)
# score_ubm = ubm.score(X_train[0])
# print(sum(score))
# print(sum(score_ubm))
# score_all = gmm.score_all(X_train[6])
# score_all_ubm = ubm.score_all(X_train[6])
# print(str(score_all) + " score_all")
# print(str(score_all_ubm) + " score_all")
# print(str(score_all/score_all_ubm) + " score_all")
end = time.time()
print(str(end - start) + " seconds")
gmm.dump(os.path.join(gmmPath, y + ".model"))
print(os.path.join(gmmPath, y + ".model"))
def load_gmm(self, label, fname):
self.y.append(label)
gmm = GMM.load(fname)
for key, val in self.kwargs.iteritems():
exec("gmm.{0} = val".format(key))
self.gmms.append(gmm)
def fit_new(self, x, label):
self.y.append(label)
gmm = GMM(self.gmm_order, **self.kwargs)
gmm.fit(x, self.ubm)
self.gmms.append(gmm)
def after_pickle(self):
self.gmms = [GMM.loads(x) for x in self.gmms]
def fit_new(self, x, label):
self.y.append(label)
gmm = GMM(self.gmm_order)
gmm.fit(x)
self.gmms.append(gmm)
def fit_new(self, x, label):
self.y.append(label)
gmm = GMM(self.gmm_order)
gmm.fit(x)
self.gmms.append(gmm)
def load_gmm(self, label, fname):
self.y.append(label)
gmm = GMM.load(fname)
for key, val in self.kwargs.iteritems():
exec("gmm.{0} = val".format(key))
self.gmms.append(gmm)
def fit_new(self, x, label):
self.y.append(label)
gmm = GMM(self.gmm_order, **self.kwargs)
gmm.fit(x, self.ubm)
self.gmms.append(gmm)
def after_pickle(self):
self.gmms = [GMM.loads(x) for x in self.gmms]
def main(datapath, modelpath, ubmpath):
fpaths = get_predict_data_fpaths(datapath)
# print "The fpath is ",fpaths
mpaths, model_name = get_model_fpath(modelpath)
# print(mpaths)
X_test, y_test, filenames = datautil.read_test_data(fpaths)
# print "X_test is",X_test
# print "Y_test is",y_test
# print(len(X_test[0]))
# return
ubm = GMM.load(ubmpath)
model = []
for mname, cname in zip(mpaths, model_name):
model.append(GMM.load(mname))
# ouputfile = open(output_filename)
resultlines = []
correct_count = 0
count_all = 0
resultdict = {}
for y in y_test:
resultdict[y] = (0, 0) # (正确数,总数)
for x, y, z in zip(X_test, y_test, filenames):
ubm_score_all = ubm.score_all(x)
score_list = []
for m in model:
score_list.append(m.score_all(x))
max_index = score_list.index(max(score_list))
print(max(score_list) / ubm_score_all)
count_all = count_all + 1
# print "The y is",y,"The Max_index is ",max_index,"The Model name is",model_name[max_index]
if y == model_name[max_index]:
correct_count = correct_count + 1
# resultstr = y + " " + model_name[max_index] + " " + z + " " + str(maxscore) + " " + str(ubm_score_all) + " " + str(maxscore / ubm_score_all)+ "\n"
resultstr = y + " " + model_name[max_index] + " " + z + "\n"
resultlines.append(resultstr)
temp = resultdict[y]
classall = temp[1] + 1
classcorrect = temp[0]
if y == model_name[max_index]:
classcorrect = classcorrect + 1
resultdict[y] = (classcorrect, classall)
resultdict["all"] = (correct_count, count_all)
print(correct_count)
print(count_all)
for k, v in resultdict.items():
print k, " ", v
output_filename_path = datapath + "/" + output_filename
outputfile = open(output_filename_path, "w")
# for line in resultlines:
outputfile.writelines(resultlines)
for k, v in resultdict.items():
outputfile.writelines(k + "\n")
for item in v:
outputfile.writelines(str(item) + "\n")
outputfile.close()
return output_filename_path
