Can we test a voice which is different from the training samples?

llSourcell / tensorflow_speech_recognition_demo

This is the code for 'How to Make a Simple Tensorflow Speech Recognizer' by @Sirajology on Youtube

383 stars 250 forks source link

Can we test a voice which is different from the training samples? #16

Open ugur45 opened 7 years ago

ugur45 commented 7 years ago

I try to use different test samples from traininig samples. For example i used this train set (0_Samantha_100.wav, 1_Samantha_100.wav, 2_Samantha_100.wav, 3_Samantha_100.wav, 4_Samantha_100.wav, 5_Samantha_100.wav, 6_Samantha_100.wav, 7_Samantha_100.wav, 8_Samantha_100.wav, 9_Samantha_100.wav), and test the 0_Samantha_120.wav file. But system not recognize test voice is "zero".

vshmyhlo commented 7 years ago

@ugur45 try to use larger training set, you have only one sample for each number which is not enough for model to generalize