Music Tech Away Day 2012

hostitchAI and Robotics

Oct 23, 2013 (4 years and 6 months ago)


Music Tech Away Day 2012

Jonathan P Wakefield

Overview of current research supervisions:

Mark Mynett

PhD p/t

Music Production for Contemporary Metal

Steve Fenton

PhD p/t

Objective Measurement of Sound Quality in Music

Braham Hughes

PhD p/t

Development of a Tool to Assist in the Creation of
Effective Audio/Visual Environments in Film and TV post

Neil Fitzpatrick

PhD p/t

The Global Studio: An evaluation of emergent
technologies and practices relating to real
time online collaboration and production
of popular music

Ruth Taylor



An Alternative Model for Enterprise in Higher Education

Chris Dewey



A Visual Interface for Mixing Popular Music

Currently being examined:

Ben Evans



The Evaluation and

of FFT and

FFT parameters for use in automatic music transcription algorithms

Interface for Mixing Popular Music

Jonathan P Wakefield

User interface for mixing (and recording)
sound really hasn’t changed in 50 years

Is it the best interface?

Does it provide all the information an
engineer needs?

Rather than the channel strip
approach can’t we provide views of
the whole mix?

We’re drawing on ideas from David
Gibson (see left) but importantly
trying to make a practical tool for

Current thinking is to flag up
problems in mix to engineer for
manual resolution but may move into
automating of aspects of mixing

Ambisonic decoders for 5.1

Jonathan P Wakefield

Development of improved

decoders for 5.1 surround sound playback

Horizontal only



2 figure of 8s)

Can convert a mono source to

B format:

Speaker feeds are then:

Ambisonic decoders for 5.1

Jonathan P Wakefield

So we need to find values 15 for
parameters for a working decoder

Determined by having a multi
objective fitness function that is based

theory of localisation which is based on:

Velocity vectors

Energy vectors

Vectors quantify low to mid and high frequency localisation performance

A heuristic search algorithm is used to find parameter sets with good fitness values

We run multiple local searches to come up with a “good” parameter set

We have also done two HPC implementations where we have ran exhaustive global
searches at reduced parameter resolution

Ambisonic decoders for 5.1

Jonathan P Wakefield

Have developed decoders that :

Provide even performance around

Take account of Human Spatial
Resolution (MMA)

Optimise for multiple off centre listeners

Optimise for constrained listening spaces
(and simultaneously determine best
speaker positions within constraints)

Also developed a Design Tool

Used Range
Removal and Importance
concepts in fitness function


Order, 2

Order, 4

Order decoders

Ambisonic decoders for 5.1

Jonathan P Wakefield

Currently doing listening tests to confirm validity of velocity and
energy vectors in determining human perception of localisation

Future Work

Trying to find a way to convert discrete 5.1 mixes into

format ?!?!?!?

Extend to with height surround sound reproduction

Compensate for playback room acoustics

Loudspeaker Arrays

Jonathan P Wakefield

Planned experiments:

Single line:

Automatically take stereo
and play across (say) 9
speakers instead of stereo

Get multi
track and pan to
9 speakers and compare to
stereo mix

Get producers to mix to
format and compare to
stereo mix

Do same for 2
D spearker

based Musical Evolution

Jonathan P Wakefield

Tim Milllea

Automated Composition of
Popular Music

Need to analyse existing good songs

Analysis feeds into a fitness function

Genetic Algorithms evolves a population
of solution guided by fitness function

GA needs a crossover function

GA also needs a mutation function

With Dr Steven Jan

Replicating the evolution of music

Unsuccessful AHRC bid this year

Supervising a Visiting Researcher/
Academic for 12 months

Also have a German
Cognitive Sciences
undergraduate coming for 3