Development of Next-Generation Music Recognition Algorithm for Content Monitoring

版权法要求音乐行业应该跟踪世界上每首歌曲和每一张唱片的每一次公开演出，以收取和准确分配版税。然而，准确的跟踪只发生在少数场景中。这个数字化转型项目将通过激发新的尖端人工智能技术，开发一种新颖的音乐识别技术（MRT）解决方案；下一代MRT算法，能够识别在现场播放或表演的音乐，或非官方的在线翻唱（例如在youtube上）。这将反过来允许词曲作者和音乐所有者获得他们应得的版税。**主要目标**开发一种使用“狭义”人工智能技术的解决方案，以识别歌曲的替代版本，而不考虑其播放风格。目前的解决方案可以很好地识别艺术家的原创音乐录音，也可以识别注册的翻唱版本。例如，只要注册了“音乐指纹”，就可以使用类似于_shazam_的解决方案来识别“Hey Jude”的录音。然而，这些现有的技术在大多数情况下无法识别音乐家和表演者在成千上万的现场音乐场所（如全国各地的酒吧、俱乐部和咖啡馆）进行的现场音乐。目标是确保现场场地和音乐表演者有一个简单的，非侵入式的自动化解决方案，以确定由于音乐所有者的版税。由于场地里的音乐没有得到认可，收藏者协会只能估计谁欠了什么；尤其是那些规模较小的音乐创作者，他们从未获得应有的版税。**主要关注领域**开发创新的MRT解决方案，以解决当前音乐识别算法无法解决的领域，即识别歌曲中版权所有者的“内容”。这个项目的动机是开发一个MVP解决方案，并定义如何将其最好地推向市场。**如何创新？**_Breathe Music的_音乐识别引擎利用人工智能开发一种阶变的、颠覆性的MRT作为识别模型，以取代现有的“录音”监控系统，如_Shazam_——能够监控和识别“录音”，也能识别录音中的“内容”。这种创新克服了只能识别和匹配以前“看到”的记录版本的限制。

**Vision**Copyright law requires that the music industry should track every public performance of every song and recording in the world to collect and accurately distribute royalties. However, accurate tracking only occurs in a minority of scenarios.This digital transformational project will develop a novel Music Recognition Technology (MRT) solution by evoking new, cutting-edge AI technologies; a next-generation MRT algorithm to enable identification of music being played or performed in a live venue, or unofficial online covers (e.g. on _YouTube_). This will in turn allow songwriters & music owners to gain the royalties due to them.**Key objectives**To develop a solution using 'narrow' AI technologies to recognise alternative versions of a song irrespective of the style in which it is being played. Current solutions are good at recognising original recordings of music by the artist and also registered recorded cover versions. For example, a recording of "Hey Jude" could be identified using _Shazam_-like solutions as long as a "music-fingerprint" of that recording has been registered. However, these existing technologies cannot in most instances identify live music undertaken by musicians & performers in the thousands of live music venues such as pubs, clubs and cafes across the land. The objective is to ensure that live venues and music performers have an easy, non-intrusive automated solution to identify the royalties due to the music owners. As a result of music in venues not getting recognised, collection societies only estimate who is owed what; the smaller music creators especially are the ones who never get the royalties they are entitled to.**Main areas of focus**On the development of an innovative MRT solution to address the areas where current music recognition algorithms are unable to address, i.e. to identify the rights-holders 'content' within the song. The motivation of this project is to develop an MVP solution and define how this could be best taken to market.**How it is innovative?**_Breathe Music's_ music recognition engine utilises AI to develop a step-change, disruptive MRT as an identification model to replace existing 'recording' monitoring systems like _Shazam_ - capable of monitoring and identifying the 'recording' but also the 'content' within the recording. This innovation overcomes the restrictions of only being able to identify and match only previously "seen" recorded versions.