2011/att11/planning_July11

We met to discuss what the next steps are after Friday's demo.

Suggestions for improving the task

  • Randomize the range (therefore they don’t know the “ninety”)?
  • New task: the nearest number to X.
  • More loudspeakers/talkers

Goal

Show effects of saliency & attention Switching effects we could apply

  • do we believe the number recognized?
  • if so, is it worth listening to?

Subcomponents…

Visual task

Simplest: LEDs that flash with the envelope, which should be unique. The visual task could

  • help to distinguish similar sounds (like p and b)?
  • direct you (saliently) to the
  • be correlated to the size of the number

Francisco Barranco has a camera he can use.

ITDs

Start with the clean signal. See if we can develop this for freefield afterwards

Input

Single-camera input Microphones or clean internal signal

Speech recognition

'Malcolm'/Kailash Patil have a novel approach using salient features.

Volunteers for various tasks

ITDs for spikes: Shih-Chii Liu

ITDs for waves: 'Merve'

Auditory recognition task: 'Merve' has started programming / 'Malcolm'/Kailash Patil have an experimental equivalent.

Getting realtime spikes to Matlab: 'Malcolm' & Shih-Chii Liu

Visual task: Francisco Barranco

Saliency & Attention modules?: 'Trevor'

Visual spikes and jAER?: 'Sam Shapero' & 'Siddarth'

Eye-tracking: Francisco Barranco (needs volunteers)

Write the wiki: ALL