Juntae, Kim

I'm really sorry for late answer, these days I'm so busy : ( . For the above questions: Maybe, your tested speech is from Aurora which contain short utterance and...

1. I cannot sure my normalization factor is perfect for every situation. However, if we use large amount of dataset which can represent the population mean and variance, the normalization...

NO, the threshold is used only in bdnn and ACAM, please refer the model definition of DNN and LSTM, their prediction is conducted by argmax function across the softmax dimension.

Hi, is there any silence in front of your sample, if not, the result may be not good. Because ACAM is context based model, there should be some samples to...

I'm very sorry for late answer. 1. audio_dir = './data/example/clean_speech.wav' # directory for specific audio file 2. the length of your audio data i.e. length(audio) in matlab notation. 3. Unfortunately...

Excellent! thank you for your interest and contributions! Because it has been a long time since I implemented it, I can't exactly remember it in detail. However, the purpose is,...

This QUESTION is quite difficult to answer, because the necessary training time is according to the training data. Therefore, you should inspect your training & validation curve to find appropriate...

1. Approximately 2 epochs for whole of my dataset I used. 2. No, that checkpoint is not the optimized one. Now, I only provide the optimized .pb file. If you...

1. If each sample file has much different noise characteristic and high noise energy, the mean and variance can be depends on noise signal rather than speech signal. However, the...

Rather than Octave, we have a plan to write our code only using the python. please wait for it thank you!