We present a variational framework for determination of intra-voxel fiber orientations from High Angular Resolution Diffusion-Weighted (HARD) MRI under the assumption of biGaussian diffusion. The approach is simultaneously estimating and regularizing the two tensor fields and the field of the proportionality corresponding to the mixture of two Gaussians. The prior information on location of the voxels with strong isotropic or one-fiber diffusion is incorporated into the energy functional in order to increase the accuracy of the estimations. The prior information on voxel classification is obtained from the spherical harmonic (SH) representation of the apparent diffusion coefficient (ADC) profiles estimated and regularized simultaneously from noisy HARD data. The performance of the proposed model has been evaluated on human HARD MR images, and the experimental results indicate the effectiveness of the model in recovering intra-voxel multi-fiber diffusion.