1.6 Birdcall identification: conclusion

Birdcall identification project

Conclusion

The combination of the matched filter process with the frequency content analysis proved to be a fairly robust tool in identifying birdcalls. The matched filter tells us which birdcall is found in the time-domain, while the frequency analysis confirms the proper frequency content at the site of the match.

Our MATLAB GUI implementation gave a simple interface to identify birdcalls within a wave file. The interface gives a visual confidence reading regarding which birdcalls are within a sound sample. With more time, the program could easily be translated to a web application or mobile interface for use in the field.

Future work

Our implementation was simply a test of concept. If we wished to take the project further, the first step would be to increase the filter library size. If made mobile, we could create a filter library that would mold itself to the species of birds in the specific geographic location, giving our algorithm less to check against.

Though our program can be made faster, the project as a whole was highly successful. Further work needs to be done to ensure the reliability of the program, especially as birdcalls are added. In the end, a venture into a marketing plan could prove to be lucrative.

Acknowledgments :

Special Thanks to:

• Our instructor, Rich Baraniuk
• Our TA, Matthew Moravec
• Our Mentor, Marcos Duarte

Sound files from:

• www.dovehunt.org
• wildspace.ec.gc.ca
• www.michiganloons.org
• www.wildernessbay.com
• www.sylvialake.org
• www.avians.net
• www.birding.com.br
• www.owling.com