System and method for encoding data in a voice recognition integrated circuit solution
作者： Yang Lin , Gao Xiang , Zhang Wenhan
DOI:
关键词:
摘要: Methods of encoding voice data for loading into an artificial intelligence (AI) integrated circuit are provided. The AI may have embedded cellular neural network implementing tasks based on the loaded data. An method generate a two-dimensional (2D) frequency-time array from audio waveform, use 2D to set arrays approximate array, load circuit, execute programming instructions contained in feed recognition result, and output result. also trains convolution (CNN) loads weights CNN tasks.
lens.org 参考文章(6)
Zafar Rafii, Audio matching based on harmonogram ,(2015)
Yu-hsin Joyce Chen, Ignacio Lopez Moreno, Sub-matrix input for neural network layers ,(2015)
Yao Song, Yu Jincheng, Combining CPU and special accelerator for implementing an artificial neural network ,(2018)
Li Xiangang, Li Chao, Method and device for extracting speech features based on artificial intelligence ,(2018)
Na Hwidong, Lee Hoshik, Speech recognizing method and apparatus ,(2018)
Tatiana Shpeisman, Balaji Vembu, Eriko Nurvitadhi, Nicolas C. Galoppo Von Borries, Altug Koker, Rajkishore Barik, Sanjeev Jahagirdar, Mark A. Anders, Sanu K. Mathew, Tsung-Han Lin, Anbang Yao, Himanshu Kaul, Vasanth Ranganathan, Abhishek R. Appu, Joydeep Ray, Kamal Sinha, Michael S. Strickland, Ping T. Tang, Xiaoming Chen, Instructions and logic to perform floating-point and integer operations for machine learning ,(2017)