Study of Senone-Based Deep Neural Network Approaches for Spoken Language Recognition