Acoustic Localization Robot

Team Members:Dave Shelley

Phil PolettiJoe Massey

Agenda

[Desired Results] [Achieved Results] [Lessons Learned] [Possible Improvements] [Recorded Demonstration] [Live Demonstration]

Desired Results

Key Desired Metrics to Achieve: Utilize quasi- three dimensional

microphone array to capture audio in real-time.

Calculate location of source utilizing beamforming algorithm

Laser points to location in space through use of Lego Mindstorm NXT through pitch and yaw movement.

Achieved Results

Key Metrics Achieved Utilizes quasi- three dimensional microphone

array to capture audio in real-time. Calculates location of source utilizing

beamforming algorithm• Instead of using a beamforming algorithm, an

amplitude vectoring algorithm was utilized • Beamforming algorithm placed overwhelmingly large

processing constraints on processor.• Did not provide far superior results as expected• Amplitude vectoring algorithm only provides a maximum

pitch and yaw resolution of 90 degrees Laser points to location in space through

use of Lego Mindstorm NXT

Lessons Learned[Issue] [Specific problems]

[Poor Quality of Purchased

Components]

USB Microphone Decided to use USB Sound Card in conjunction with 3.5mm analog

microphone•Poor Frequency Response

•Inconsistent gain across different mics

[Undesired Environment

Noise]

Acoustical Inconsistencies Shape of room causes undesired acoustic reflections Electronics utilized for project implementation cause undesired interference

[Beamforming Algorithm]

Algorithm causes large system latencies Complexity of algorithm makes implementation difficult Originally intended for use with 100+ microphones over a large spatial

region

[USB Interfacing Problems]

Operating System Permissions USB Hub issues

Possible Improvement

[Improvement] [Best method for implementation]

[TDOA (Time Delay of Arrival)]

GCC –PHAT (Cross Correlation with Phase Transform) would provide much higher resolution, detecting phase differences, rather than amplitude differences between microphones.

Additional Microphones]

Currently, the amplitude vectoring algorithm only allows for 90 degrees maximum of resolution.

More microphones in the X and Y planes would allow for much higher resolution.

[Threading] Utilize threading of all processes to streamline

[Error Checking] Compare five or more concurrent runs and determine location based on histogram of five runs

[Laser Driver] Implement a voltage regular to more easily pulse the laser without using Lego NXT Motor Drivers.

For Further Information:

For a complete explanation of our project as

well as compilations of all our code,

proceed to:

http://code.google.com/p/acoustic-localization-robot/

Hardware Components Utilized

Dell Inspiron

2.0 GHz Core2Duo

800MHz Front Side Bus

2 GB DDR2 RAM

Linux Ubuntu v9.10

6x USB Microphone Array Lego NXT

Microcontroller with Laser Turret

Breakdown of Debug Screen

Band Pass Filter Coefficients

Amplitude in dB of highest frequency

on each mic

Location of Max Amplitude

Correct USB Handshake to Lego

NXT

Recorded Demonstration

Shortcut to Recorded Demonstration

Live Demonstration