- numpy.cumsum():按照所给定的轴参数返回元素的梯形累计和,axis=0,按照行累加。axis=1,按照列累加。axis不给定具体值,就把numpy数组当成一个一维数组。
- np.pad():常用于深度学习中的数据预处理,可以将numpy数组按指定的方法填充成指定的形状
如下,在 speech emotion recognition中,将不满300帧的特征矩阵用np.pad补灵
if(time <= 300):
part = mel_spec
delta11 = delta1
delta21 = delta2
part = np.pad(part,((0,300 - part.shape[0]),(0,0)),'constant',constant_values = 0)
delta11 = np.pad(delta11,((0,300 - delta11.shape[0]),(0,0)),'constant',constant_values = 0)
delta21 = np.pad(delta21,((0,300 - delta21.shape[0]),(0,0)),'constant',constant_values = 0)
traindata1[train_num*300:(train_num+1)*300] = part
traindata2[train_num*300:(train_num+1)*300] = delta11
traindata3[train_num*300:(train_num+1)*300] = delta21