A Speaker Localization System for Lecture Room Environment
This paper presents a speaker localization system, which is an entry to Rich Transcription 2005 Spring Meeting Recognition Evaluation. The system is developed in the Institute of Signal Processing at Tampere University of Technology (TUT). The paper describes the framework of the evaluation and the proposed localization system. This paper is an extension to  giving the actual performance values of the system.
The localization system is based on spatially separate sensor stations. The sensor stations estimate Direction of Arrival (DOA) of acoustic wavefronts. Each sensor station produces a three dimensional DOA vector. The estimated DOA vectors at each time instant are combined to calculate the location of the sound sound source.
The performance of the system was determined using a set of predefined metrics. Using multiple metrics enables one to evaluate the performance of the localization system from different viewpoints. The overall performance is characterized by RMS error between estimates and reference positions. The results show that the performance of the proposed system is consistent and accuracy is satisfactory for meeting room scenario. However, several improvements can be seen.
KeywordsLocalization System Sound Source Sensor Station Localization Task Stream Segregation
Unable to display preview. Download preview PDF.
- 1.Pirinen, T., Pertilä, P., Parviainen, M.: The TUT 2005 source localization system. In: Rich Transcription 2005 Sring Meeting Recognition Evaluation, July 13, 2005. Royal College of Physicians, Edinburgh (2005)Google Scholar
- 3.Bregman, A.S.: Auditory Scene Analysis. The MIT Press, Cambridge (1990)Google Scholar
- 4.National Institute of Standards and Technology: Spring 2005 (RT-05S) Rich Transcription Meeting Recognition Evaluation Plan. (2005), http://www.nist.gov/speech/tests/rt/rt2005/spring/rt05s-meeting-eval-plan-V1.pdf
- 5.Omologo, M., Brutti, A., Svaizer, P.: Speaker localization and tracking – evaluation criteria. Technical report, National Institute of Standards and Technology (2005), http://www.nist.gov/speech/tests/rt/rt2005/spring/sloc/CHILIRST_SpeakerLocEval-V5.0-2005-01-18.pdf1
- 6.Surcin, S., Stiefelhagen, R., McDonough, J.: D7.4 evaluation packages for the first CHIL evaluation campaign. Technical report, Computers in the Human Interaction Loop (CHIL) Consortium (2005)Google Scholar
- 7.Stiefelhagen, R.: CHIL evaluation data – overview of sensor setup and recordings. Technical report, Computers in the Human Interaction Loop (CHIL) Consortium (2004)Google Scholar
- 9.Yli-Hietanen, J., Kalliojärvi, K., Astola, J.: Low-complexity angle of arrival estimation of wideband signals using small arrays. In: Proceedings of the 8th IEEE Signal Processing Workshop on Statistical Signal and Array Signal Processing, pp. 109–112 (1996)Google Scholar
- 11.Pirinen, T.: Normalized confidence factors for robust direction of arrival estimation. In: Proceedings of the 2005 IEEE International Symposium on Circuits and Systems (ISCAS) (2004)Google Scholar
- 12.Kaplan, L., Le, Q., Molnár, P.: Maximum likelihood methods for bearings-only target localization. In: Proceedings of the 2001 IEEE International Conference on Acoustics Speech, and Signal Processing (ICASSP 2001), pp. 3001–3004 (2001)Google Scholar
- 14.Pertilä, P., Parviainen, M., Korhonen, T., Visa, A.: A spatiotemporal approach to passive sound source localization. In: International Symposium on Communications and Information Technologies (ISCIT 2004) (2004)Google Scholar