comparison Code/genre_classification/classification/preprocess_spectrograms_7digital.py @ 24:68a62ca32441

Organized python scripts
author Paulo Chiliguano <p.e.chiilguano@se14.qmul.ac.uk>
date Sat, 15 Aug 2015 19:16:17 +0100
parents
children
comparison
equal deleted inserted replaced
23:45e6f85d0ba4 24:68a62ca32441
1 # -*- coding: utf-8 -*-
2 """
3 Created on Thu Jul 23 21:55:58 2015
4
5 @author: paulochiliguano
6 """
7
8
9 import tables
10 import numpy as np
11 import cPickle
12 import sklearn.preprocessing as preprocessing
13
14 #Read HDF5 file that contains log-mel spectrograms
15 filename = '/homes/pchilguano/msc_project/dataset/7digital/features/\
16 feats.h5'
17 with tables.openFile(filename, 'r') as f:
18 features = f.root.x.read()
19 #filenames = f.root.filenames.read()
20
21 #Pre-processing of spectrograms mean=0 and std=1
22 n_per_example = np.prod(features.shape[1:-1])
23 number_of_features = features.shape[-1]
24 flat_data = features.view()
25 flat_data.shape = (-1, number_of_features)
26 scaler = preprocessing.StandardScaler().fit(flat_data)
27 flat_data = scaler.transform(flat_data)
28 flat_data.shape = (features.shape[0], -1)
29
30 f = file('/homes/pchilguano/msc_project/dataset/7digital/features/\
31 feats.pkl', 'wb')
32 cPickle.dump(flat_data, f, protocol=cPickle.HIGHEST_PROTOCOL)
33 f.close()