Manifold learning for user profiling and identity verification using motion sensors: CNN-designed architecture and models

The table Designed CNN Architecture describes the final designed CNN model for the proposed method, which was inspired in the MobileNet architecture. The number of parameters of each layer and the total of this model are specified in this table.
The table Convolution Blocks contains details about the different convolution layers adopted in the architecture.
We also made available the weights of both trained models: Spectrogram model and ADT model.

表格《设计卷积神经网络架构（Designed CNN Architecture）》详细介绍了本文所提方法的最终卷积神经网络（Convolutional Neural Network，CNN）模型，该模型的设计灵感源自MobileNet架构。本表格同时列明了模型各层的参数量与模型总参数量。
表格《卷积模块（Convolution Blocks）》详细记载了该架构中采用的各类卷积层的相关细节。
本次研究还公开了两个已训练完成模型的权重参数：语谱图模型（Spectrogram model）与ADT模型。