In this paper, we present a Brain Computer Interface (BCI) system using multichannel functional near-infrared spectroscopy (fNIRS) signal acquired when subjects execute speed and force imagination of right hand. Our goal is to classify much more movement imagination details so that a BCI system can provide more control commands, which is helpful for BCI application. Six subjects (3 male, 3 female) participate in the experiment for 3 sessions. We use Gaussian filter and wavelet-MDL to preprocess the acquired signal, and then use support vector machine (SVM) to classify task state versus rest state and speed imagination versus force imagination. Our results show that using oxyhemoglobin (HbO) data as feature can get comparable results with condition of using both HbO and hemoglobin (Hb) data as feature. Also, feature from left head provide more information than right head when subjects doing right hand movement imagination. Finally we study feature period effects on classification accuracy and find some key periods dominating the results. Our study demonstrates that an fNIRS based BCI have the potential to provide more control commands for a real application.