Structural sparsification for Far-field Speaker Recognition with GNA
This work addresses the need for efficient speaker recognition on mobile hardware, though it is incremental as it builds on existing methods.
The paper tackles the problem of high computational cost in far-field speaker recognition on mobile devices by applying structural sparsification to time-delay neural networks, resulting in a model that removes 60% of parameters, increases equal error rate by only 0.18%, and achieves over 1.5x speedup.
Recently, deep neural networks (DNN) have been widely used in speaker recognition area. In order to achieve fast response time and high accuracy, the requirements for hardware resources increase rapidly. However, as the speaker recognition application is often implemented on mobile devices, it is necessary to maintain a low computational cost while keeping high accuracy in far-field condition. In this paper, we apply structural sparsification on time-delay neural networks (TDNN) to remove redundant structures and accelerate the execution. On our targeted hardware, our model can remove 60% of parameters and only slightly increasing equal error rate (EER) by 0.18% while our structural sparse model can achieve more than 1.5x speedup.