Speech-driven 3D facial animation has achieved significant progress, producing increasingly realistic results. However, animations generated solely from audio input often lack expressive style. Current research predominantly focuses on style control using video or mesh sequences, neglecting scenarios relying exclusively on audio input, which results in less expressive outcomes. To address this limitation, we propose a novel method that employs a regional attention mechanism guided by a single identity vector, enabling the generation of stylistically consistent and highly expressive animations. Our approach customizes facial attention mask regions for individual speakers and incorporates two loss functions, \(L_{mask}\) and \(L_{normal}\) , to capture detailed style information through vertex movements and geometric variations. Extensive experiments and user studies demonstrate that our method outperforms existing approaches in qualitative and quantitative evaluations.