隨著科技日新月異，人們之間的距離因此縮短，文化的交流讓世界成為地球村，促使東西方的文化融合更加快速。在現代人的消遣娛樂當中，自從臺灣華語流行歌曲男歌手周杰倫與歌詞創作家方文山合力創作出《娘子》、《東風破》、《菊花台》等歌曲後，受到熱烈的回響，娛樂產業便刮起了一陣「中國風」。由於中國風的興起，更出現了西方樂曲被翻唱成中國風格的作品，翻唱（Cover）一直是樂壇中盛行之事，是由另外一位作者重新詮釋原作音樂，此種作為不僅限於歌唱，亦適用於跨樂器的詮釋，俗稱翻奏。 此篇論文，主要是提出一個基於圖像人工智慧的技術，來成就更好的音樂風格轉換，並且搭上時下流行的中國風，將音樂轉換成中國古箏的版本。透過人類感知實驗評估此方式的結果，獲得4.3的高分（0分為最低分，5分為最高分）。 ;Chinese Guzheng music is popular from ancient Qin dynasty. However, from the study of modern Chinese history, Chinese people continued to pursue westernization. The Chinese Guzheng also faces the drastic changes of the external environment such as the needs of modern national orchestras. Due to changes, the Guzheng is reformed in the direction of expanding the sound range, increasing the volume. Although new design for the Chinese Guzheng instrument has been proposed, it is too hard for players to catch the music transition because of long time producing and learning new skills. To improve this issue, we propose music conversion with CycleGAN which is a deep learning technique that involves the automatic training of image-to-image translation models without paired examples. Moreover, this method provides a simple way which can be used by whom never learn any instrument. This is a brand new research with good results.
