I-Corps: Automatic Music Generation Using Artificial Intelligence and Music Theory

这个I-Corps项目的更广泛的影响/商业潜力是彻底改变音乐产业和相关部门，如游戏开发和视频制作，这些都与音乐有关。利用所提出的音乐生成技术，可以避免版权问题，因为所生成的音乐是完全原创的。此外，与传统的作曲方法相比，用户或作曲家至少需要一个小时才能制作一分钟的音乐作品，所提出的技术只需不到一分钟即可生成相同长度的音乐。这些优势将使小型游戏开发者和自媒体制作者能够在很短的时间内以最小的成本制作出符合他们需求的原创音乐，这对游戏和自媒体行业都有很大的好处。I-Corps项目的特色是基于人工智能和音乐理论的用户哼唱自动生成音乐技术。没有任何专业的音乐知识，用户可以简单地从哼一小段旋律开始。将AI算法与音乐理论相结合，应用程序完成其他必要的音乐组件（如和弦进行，节拍和克里思），并将短输入旋律扩展为完整的音乐作品，用户可以更改流派并进行后期编辑。所提出的技术的技术差异有三个方面。首先，一个音高侦测演算法被发展来辨识使用者哼唱的曲调，并将其转换成音符，然后将其优化以符合音乐理论与逻辑。其次，结合递归神经网络和音乐理论的一种新的算法被开发出来，生成一个完整的音乐作品，包括主旋律，和弦进行，动机和伴奏，基于用户的短输入旋律。第三，后期编辑阶段允许用户根据生成的和弦进行修改音符和流派。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is to revolutionize the music industry and related sectors such as game development and video production which significantly involve music. With the proposed music generation technology, copyright issues can be avoided since the music generated is completely original. Furthermore, compared to traditional composing methods which take at least an hour for the user or composer to produce a one-minute piece of music, the proposed technology spends only less than a minute to generate the music of the same length. These advantages will allow small game developers and self-media producers to generate original music that fit to their needs in a very short amount of time with minimum cost, which significantly benefits the game and self-media industries.This I-Corps project features the technology of automatic music generation from user's humming based on AI and music theory. Without any professional knowledge of music, a user can simply start by humming a short melody. Combining the AI algorithm with music theory, the application completes other necessary musical components (such as chord progression, beats and tempo) and expands the short input melody into a full piece of music for which the user can change genre and do post-editing. The technical differentiators of the proposed technology are threefold. Firstly, a pitch detection algorithm is developed to recognize the user's humming tune and turn it into musical notes, which are then optimized to conform to music theory and logic. Secondly, a novel algorithm combining Recurrent Neural Network and music theory is developed to generate a full piece of music, including the main melody, chord progression, motive and accompaniment, based on the user's short input melody. Thirdly, the post editing stage allows the user to modify the notes and genre according to the generated chord progression.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.