Music and machine learning
It is no secret that music affects our mood. Different songs can put us in a good mood, others lead to tears, and yet others calm down. There is a reason for them in movies or commercials amplifier communication. Recently, scientists from the University of Southern California decided to check what determines in music that it works in a specific way. Of course, it is difficult to approach such a topic without adequate facilities. Due to the fact that people have a problem with noticing all the nuances, it was decided to use machine learning algorithms.
Also read: What is transhumanism?
It began with collecting material for research. Scientists needed songs – sad and cheerful. To this end, they searched internet music sites and discussion forums to find songs tagged "sad" and "cheerful" or their synonyms. They wanted to reduce the likelihood that the study participants knew the songs before, so they decided to pick out niche works with few plays from the web.
This created a playlist with 120 songs. Their fragments were given to eight people for interrogation, whose task was to assess what emotions they feel in connection with the music they hear. In the case of 27 songs, at least 6 out of 8 people agreed on these matters – these songs were then sent to the online survey. Then the songs were presented to a group of one hundred people. Some participants were connected to brain activity equipment at that time, while others were testing their heart rate.
Also check: Comparison of the best streaming services.
The songs were checked for 74 different properties, such as dynamics, harmony and rhythm. The task of machine learning algorithms was to find a relationship between the content of the work and the physiological response of the body, and more precisely – to determine which of several dozen properties should be observed to predict the body's response. For example, it turned out that the timbre (i.e. the intensity of medium and high frequencies), volume and clarity of the rhythm in sad songs affect brain activity. In turn, the color, complexity, rhythm clarity and tone predictability are correlated with heart rate changes.
Modern art and medicine
We now move on to how this research can be used. Namely, scientists want to create appropriate machine learning models that will be able to answer the question what exactly emotions will evoke, how they will affect our psyche. It certainly sounds fascinating and frightening at the same time.
Also check: Artificial intelligence in the service of our health.
Currently, it is said that ideally adapted music for listeners would work in psychotherapy. Undoubtedly, this would be a great example of how machine learning can literally help us. I can see, however, that we will soon see such personalized songs in commercials or music streaming services, such as Spotify or Apple Music.
source: Artificial intelligence
More from category Science: