Share this content

Listening to the mood music with emotion analytics

2nd Jul 2018
Share this content

Like many geeky teenagers in the 80's, I was fascinated by electronic music and especially bands like Kraftwerk that were developing the technology as well as the genre, and I ended up building a synthesizer in the vain hope of releasing my inner electro pop star; suffice to say that knowledge without talent (and practice) is not enough to top the charts (my teens predated the punk era).

One of the things I did learn was about sound 'envelopes'; specifically: Attack, Decay, Sustain and Release (Also known as ADSR). 

In a synthesizer, ADSR controls and modifies the amplitude (volume) of a sound during 4 distinct phases; after the initial stimulus (e.g. key press - attack), whilst the key remains pressed (decay and sustain) and what happens when the key is let go (release). It partly explains why a harp does not sound like a piano, even though both use strings. However, I squirrelled this knowledge away with no expectation of ever finding a use for it again.

ADSR Sound Envelope

So why think about it now? because I am working a lot with emotions and how to analyse and represent them in predictive computer models of behaviour.

Early in my research, I realised that I was going to have to find a way to take into account that emotions are not simply 'on or off' or maintain a continuous amplitude once the stimulus has passed. In some ways, the intensity of emotions is very like an ADSR envelope - with different envelopes for different emotions and in response to different stimuli (e.g. like the difference between striking a string and plucking it, a piano and a harp).

  • Attack phase - how quickly the intensity of an emotion changes in response to the initial stimulus;
  • Decay phase - how quickly it starts to subside, even though the stimulus remains in place;
  • Sustain phase - the lingering effect of a continuous stimulus; and finally
  • Release phase - the rate of return to an underlying normal state once the stimulus is removed.

Taking two core emotions as examples - the intensity envelope for surprise looks very different to the one for joy.

Emotions are not simply 'on or off' - they are far more complex and subtle than that.

To continue the analogy; the complex interactions of emotions is a little like playing chords - some pleasing, some discordant and jarring. Repeated stimuli can add to the overall sounds or distract. I could go on but you probably get the picture (or hear the tune) by now.

In my opinion, it is not enough to say that a customer feels 'happy' or 'sad', or even 'very happy' or a 'little sad' - if we are to get even close to analysing and modelling emotions over time (and I have), we are going to have to start listening to the music - especially if we ever hope to be able to play.

Replies (0)

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.