AI automatically filters matching music to brands
New standard tool
The novel music recommendation system is based on accumulated expert knowledge of music effects research, empirical findings from a multinational listening experiment and an algorithm-based analysis procedure, and for the first time represents a scientifically validated decision-making aid for the selection of suitable music tracks for individual brands. "The ABC_DJ algorithm can predict the fit of music to brand with an accuracy of 80.1%," says Dr. Jochen Steffens of TU Berlin.The powerful algorithm draws exclusively on the audio content of the songs themselves and not on manually assigned keywords and descriptions. With this software, brands and advertising agencies can automatically and accurately find the right music for any brand or campaign. The target group model "Sinus-Meta-Milieus®" was also integrated into the ABC_System so that the algorithm provides brand-appropriate music for target groups from all relevant socio-cultural milieus and thus delivers appropriate brand music for every context. Robin Hofmann, Co-Founder and Creative Director at HearDis! concludes, "The ABC_DJ process can now be considered a standard tool that creative agencies use to describe brands and brand music."
Research participants describe music properties
To create such a system, the researchers first developed a vocabulary to systematically describe music in a branding context. This novel "General Music Branding Inventory" was created with marketing experts. Subsequent analyses then used a pool of 28,543 songs, 549 of which were subjected to detailed evaluation. A large-scale listening experiment with 10,144 participants in Germany, Spain and the UK yielded 53,344 measurements based on 2,018,704 collected data points. The sample was balanced in terms of age, country, and education to ensure a representative insight into the perception of semantic expression in music by different target groups.
Powerful algorithm
Paris-based ABC DJ project partner IRCAM (Institute for Research and Coordination in Acoustics / Music) calculated a variety of audio features for each of the 549 songs, which included information about sound, harmony, rhythm, instrumentation, genre and style. Highly effective machine learning techniques were then used to find the features best suited to predict people's music ratings.But how exactly does the recommendation algorithm work? It is first based on four basic factors called emotional valence, emotional arousal, authenticity, and recency. Although target groups verbally describe a given piece of music in different ways, it is generally possible to distill and harmonize their descriptions using the above factors: A given piece can be described as more or less joyful (emotional valence), intense (emotional arousal), authentic, and progressive.Please click here to listen to a music excerpt predicted by the algorithm to be bright, playful and fun:Please click here to listen to a music excerpt that the algorithm describes as loving, kind andwarm was predicted:
Music perception
Looking into the heart of the ABC DJ algorithm, one finds a deep insight into how music is perceived and how it moves the listener. Musical characteristics such as rhythm and dynamics, as well as instrumentation (e.g., piano, guitar, vocals) and musical styles (e.g., classic rock, deep house) are crucial to perceived musical expression.For example, the results of the listening experiment show that people are more likely to perceive music with a high dynamic range, especially pieces belonging to the Historical Classic, Tango, Flamenco, and Blues styles, which are considered more authentic than other genres. In contrast, pieces of electronic music, especially those with low microdynamics and heavy sub-bass frequencies, are perceived as less authentic. The researchers further found that the louder, faster, and sharper sounding or more discordant a song is, the more emotionally authentic it is perceived to be. Finally, songs from the styles of electro, EDM, pop, rock and R'n''B are perceived as very progressive, while German Schlager is seen as more traditional.
Audio branding in real time
The results of ABC-DJ represent a counter-model to the "one-sound-for-all" formula currently offered by many audio branding companies and in-store music providers. Because ABC-DJ nundis the ability to extract brand-appropriate, musical properties directly from audio signals, new music can be discovered from any digital music library in real time and added to the music pool for a particular brand.This will greatly expand the range of music that can be used by creative agencies and brands. This will also benefit artists and independent record labels, for whom ABC-DJ-based audio branding will open up new means of monetizing music, for example by using it as background music in commercials or as so-called in-store music in stores and restaurants.Robin Hofmann concludes, "ABC_DJ aims to strengthen the European creative industries in global competition. With this innovation project, we want to equip players in acoustic brand management with innovative software tools and make them independent of US players who dominate the market with North American chart music. It's also about European cultural identity and access to new business models for European artists."
Europe-wide audio branding project
The ABC_DJ project was funded under the European Commission's Horizon 2020 grant program (No. 688122) and involves seven companies and research institutions fromfive different European countries. The overarching goal of the three-year project is to provide European creative agencies with sophisticated software tools that support the entire process of music creation, intensively promote audio branding campaigns, and support the European business location.About HearDis!HearDis! is a full service agency for corporate sound based in Berlin and Stuttgart. HearDis! has been active since 2005 and currently employs 21 people. HearDis! GmbH takes on the conception, planning and implementation of acoustic brand communication. The music consultants support brands in the areas of audio branding, in-store music, voice & music production, music consulting & licensing, and content marketing with music.
