Joint Voice Identification Separation Research

Current Research


Audio Feature Comparison


Extracted audio features have different characteristics, which differentiate the quality of each in classification methods. Here the features are compared through analysis and synthesis of audio examples. This is done by first extracting acoustic features from speech and then reconstructing the speech based on the extracted features.

Single Speaker Identification


Presented Posters


Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA) 2007 Poster


This research examines the "cocktail party" problem which is generally described as the brain's ability to differentiate between multiple sounds that occur at the same time. Our system follows this idea by iteratively attempting to estimate the speech coming from each person’s mouth in a mixture of voices and identify those individuals speaking in the room. Through iterations the system attempts to gain more information about the speakers for better estimation and identification performance.

Drexel University Research Day 2007


The voice contains many distinguishing characteristics that enable a person’s speech to be implemented as a method of identification. This research investigates the performance of the voice as a biometric identifier in single speaker and two speaker instantaneous mixture configurations. The performance of both configurations were tested using both Kullback-Leibler divergence and Neyman-Pearson detection classification methods.


Publications


  • Kim, Y. E., Walsh, J. M., and Doll, T. M. (2008). Comparison of a joint iterative method for multiple speaker identification with sequential blind source separation and speaker identification. Proceedings of the 2008 IEEE Odyssey Workshop on Speaker and Language Recognition, Stellenbosch, South Africa: IEEE. [PDF]

  • Walsh, J. M., Kim, Y. E., and Doll, T. M. (2007). Joint iterative multi-speaker identification and source separation using expectation propagation. Proceedings of the 2007 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, New Paltz, NY: IEEE. [PDF]