*Graphic interpretation of the most common clustering techniques**Example of a passive aperture response to different incoming signals**ICSI Speaker Diarization for Broadcast News blocks diagram**Acoustic models for speaker clustering**Speaker turn duration histograms**Overlap histograms in RT06s conference room meetings**Main blocks involved in the meetings speaker diarization system**single-channel speaker diarization for meetings block diagram**Speaker models initialization based on Gaussian splitting**Cross-validation EM training algorithm**Energy-based detector blocks diagram**Left, filter over . Decision of silence in red after the thresholding.**State machine used to apply time constraints.**Hybrid Speech/non-speech detector blocks diagram**Clusters initialization blocks diagram**Friends-and-enemies clusters initialization process**Cluster models with Minimum duration and modified probabilities**Possible Speaker clustering errors due to clusters purity problems**Speech-silence histogram for a full meeting**Observed assignment of frames to Gaussian mixtures**Evaluation of metric 1 on two clusters given their models**Speech/non-speech histograms for different possible model complexities**Linear microphone array with all microphones equidistant at distance d**Filter and sum algorithm blocks diagram**filter-and-sum implementation blocks diagram**Cross-correlation values histograms for RT06s AMI meetings**Filter and Sum double-Viterbi delays selection**Two-step TDOA Viterbi decoding example, step 1**Two-step TDOA Viterbi decoding example, step 1 for an individual channel**Two-step TDOA Viterbi decoding example, step 2**Multichannel delayed signal sum using a triangular window**Locations information contained in the TDOA values**Fusion of TDOA values and acoustic features within the speaker diarization module**First merge cluster-pair BIC values and histogram for acoustic and TDOA features**Acoustic weight evolution with the number of iterations for meeting CMU_20050912-0900**Energy-based system errors depending on its segment minimum duration**Model-based system errors depending on its segment minimum duration**Individual meetings DER vs. SNR vs. number of microphones in the RT06s system**Development set SNR modifying the percentage of noise threshold adjustment**Development set SNR values modifying the Viterbi transition prob. weights in the F&S algorithm**Development set SNR values modifying the number of N-best values used for TDOA selection**DER for the model complexity selection algorithm using different CCR values**DER for the initial number of clusters algorithm using different CCR values**DER for the combination of complexity selection + initial number of clusters using different CCR values**DER variation with the number of parallel models used in CV-EM training**DER variation with the number of friends used in the friends-and-enemies initialization**DER variation with the percentage of accepted frames and used Gaussians in frame purification**DER scores for the baseline system setting the relative weights by hand on development data**DER evolution with the weight computation iterations**DER evolution changing the initial feature stream weights**DER variation with the number of Gaussian mixtures initially assigned to the TDOA models**DER variation with the CCR parameter in the agglomerate system**DER variation with the number of friends in the agglomerate system**DER variation with the number of EM iterations of a standard EM-ML training algorithm**DER variation with the number CV-EM parallel models**DER variation with the frame % acceptance for frame purification algorithm**DER variation with the Gaussian % used in the frame purification algorithm**DER Break-down by meeting for the RT05s conference data**DER break-down by show for the RT05s lecture data**DER break-down by show for the RT06s conference data**DER break-down by show for the RT06s lecture data*

user 2008-12-08