Abstract
Quantifying the pre-trained neural network model can reduce its storage size and speed up the forward-inference process. Previous researches for the quantization of voice wake up networks couldn’t reach our anticipation for applying it to applications. The voice wake up network is mainly composed of recurrent neural network and classification function. Recurrent neural network can take preorder words into consideration and it widely used in voice network. In this paper, we propose method to quantize the structure of recurrent neural networ including RNN, GRU and LSTM by optimizing the method of quantifying the activation function. At the same time, using the translation invariance of the classification function to improve classification accuracy on Speech commands dataset version 2. Using the method proposed in this paper, we can make the quantized model achieve a better voice wake up rate.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
