Minha apresentação de 1 minuto no Poster Craze Session 1
- One of the interests of our group is real-time Digital Signal Processing.
- In this research project, we are interested in performing DSP on highly
available devices in terms of technology and cost.
- Other platforms we are studying are GPU cards and Arduino boards, but this
work is specifically about evaluating the performance of Android devices for
real-time DSP.
- So the initial question was how to obtain useful information/statisctics
about Android devices performance for DSP which can aid aesthetic and
algorithmic decisions regarding hardware (model) and software (version)
choices.
- We developed a minimal Android application for DSP which can read from the
microphone or from WAV files stored on the memory and perform a series of
common DSP techniques, as for example FFTs, recursive filtering, additive
synthesis, etc.
- This application includes timekeeping code, so we can keep track of the
amount of time required to perform each of these tasks while maintaining
real-time feasibility.
- This allows for giving liver feedback to the musician/developer about
resources consumption and device performance, and also generating reports
that can be further analised.
- With this data, we can obtain useful information about devices as for
example how long they take to perform an N point FFT and if this is indeed
feasible under real-time constraints.
- We can also obtain the maximum amount of FIR filter coefficients that can be
applied to the input while maintaining real-time conditions.