Optimised GSVD for Dysphonic Voice Quality Enhancement