Meena Ramani, Kaustubh Kale, Dr John G Harris

1
Direction of Arrival (DOA) estimation algorithm

• Meena Ramani, Kaustubh Kale, Dr John G Harris

2
DOA talk outline

• Motivation
• DOA algorithm requirements
• Types of DOA algorithms
• Accuracy vs. Computational intensity

3
Where does the DOA come into the picture?
7 is good for me too!!
DOA Estimation qs qn
Lets meet at 11?!?
11 sounds good!!
Beamformer
Has 2 microphones
4
DOA requirements

• The DOA algorithm must satisfy the following
  conditions
• High accuracy particularly at low SNRs
• Low computational intensity
• Low memory
• Efficient for a 2 microphone array with 4cm
  separation (set by beamforming requirements).

5
DOA methods
6
DOA Method Equation for Implementation
Delay and Sum
Minimum Variance
MUSIC
Coherent MUSIC
Root MUSIC
ESPRIT
7
Comparison of accuracy at different SNR values
Estimated DOA over a sentence
Estimated Speech DOA over a sentence
RMSE at different SNRs
8
Evaluation of computational IntensityMFLOPS
comparison chart
Min Variance 0.93 Mflops Coherent MUSIC 0.39
Mflops DS 0.35 Mflops MUSIC 0.081 Mflops
ESPRIT 0.0086 Mflops
Root MUSIC 0.0068 Mflops
9
Comparison of Accuracy-MFLOPS
10
Conclusion

• Tradeoff between Accuracy and MFLOPS leads to the
  conclusion that ESPRIT is probably the DOA
  algorithm best suited for our purpose
• MFLOPS value 0.0086
• RMSE value3o (at 10dB)

Thank you.
Questions?
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)