中文聲調在語言學上，已被確定與語詞和句子的辨識度有關。但是在神經心理學方面，有關聲調產生的大腦功能區仍有爭議性。因此本研究的目的是要用類神經網路來研究中文語音聲調產生之神經關聯性。本研究改良以類神經網路理論為基礎的模型-DIVA模型(Directions Into Velocities Articulator)做為工具，使DIVA模型可以產生不同聲調的中文語音並產生模擬的大腦活化區。模擬結果（相同聲調不同母音（/u/－/a/）和不同母音相同聲調之學習（/i/－/a/））顯示改良後的模型可以產生不同的聲調以及維持原本的功能。進一步模擬產生相同母音不同聲調時（ㄚ、ㄚˊ、ㄚˇ、ㄚˋ）的結果顯示在DIVA設定的運動區與運動前區中，喉嚨的位置以及感覺區會有不同的活化顯示。這些活化區和臨床實驗結果類似，顯示改良後的DIVA模型具有探討中文聲調之神經關聯的功能。 In linguistics, Chinese tone is related to the Chinese word and sentence recognition rates. Neuropsychologically, issues of neural correlates on Chinese tone are still inconclusive. The purpose of this study is to investigate neural correlates on Chinese tone using a Neural Network Model – DIVA model ( Directions Into Velocities Articulator) . This study modified the original DIVA model to include additional tone production to simulate Chinese tone production and the resulted brain activation. Simulated results (different vowels with the same tone ( /a/ - /u/ ) and training process of different vowels with the same tone (/a/ - /i/)) verified new functions of the modified DIVA model. Further simulation on the same vowel with four different Chinese tones production(ㄚ、ㄚˊ、ㄚˇ、ㄚˋ)were implemented and compared to the results of published clinical studies. Similar brain activation areas were found on the larynx area in the motor cortex, pre-motor, and somatosensory cortices set by the DIVA model. These simulated results show that the improved DIVA model could be used to study the neural correlates on Chinese tone production.
