Search

In the recent years, extensive researches have been performed on various possible implementations of speaker diarization systems. These systems require efficient clustering algorithms in order to improve their performances in real-time processing. Teaching–learning-based optimization (TLBO) is such clustering algorithm which can be used to resolve the problem to the optimum clustering in a reasonable time. In this paper, a real-time implementation of speaker diarization (SD) system on raspberry pi 3 (RPi 3) using TLBO technique as classifier has been performed. This system has been evaluated on broadcasting radio dataset (NDTV), and the experimental tests have shown that this technique has succeeded to achieve acceptable performances in terms of diarization error rate (DER = 21.90% and 35% in single- and cross-show diarization, respectively), accuracy (87.30%), and real-time factor (RTF = 2.40). Also, we have tested TLBO technique on a 2.4 GHz Intel Core i5 processor using REPERE corpus. Thus, ameliorated results have been obtained in terms of execution time (xRT) and DER in both tasks of single- and cross-show speaker diarization (0.08 and 0.095, and 18.50% and 26.30%, respectively).

International audience; State of the art diarization systems now achieve decent performance but those performances are often not good enough to deploy them without any human supervision. In this paper we propose a framework that solicits a human in the loop to correct the clustering by answering simple questions. After defining the nature of the questions, we propose an algorithm to list those questions and two stopping criteria that are necessary to limit the work load on the human in the loop. Experiments performed on the ALLIES dataset show that a limited interaction with a human expert can lead to considerable improvement of up to 36.5% relative diarization error rate (DER) compared to a strong baseline.

Speaker clustering is a crucial step for speaker diarization. The short duration of speech segments in telephone speech dialogue and the absence of prior information on the number of clusters dramatically increase the difficulty of this problem in diarizing spontaneous telephone speech conversations. We propose a simple iterative Mean Shift algorithm based on the cosine distance to perform speaker clustering under these conditions. Two variants of the cosine distance Mean Shift are compared in an exhaustive practical study. We report state of the art results as measured by the Diarization Error Rate and the Number of Detected Speakers on the LDC CallHome telephone corpus.

In this paper, we present a new approach for speaker diarization. First, we use the prosodic information calculated on the original speech to resynthesize the new speech data utilizing the spectrum modeling technique. The resynthesized data is modeled with sinusoids based on pitch, vibration amplitude, and phase bias. Then, we use the resynthesized speech data to extract cepstral features and integrate them with the cepstral features from original speech for speaker diarization. At last, we show how the two streams of cepstral features can be combined to improve the robustness of speaker diarization. Experiments carried out on the standardized datasets (the US National Institute of Standards and Technology Rich Transcription 04-S multiple distant microphone conditions) show a significant improvement in diarization error rate compared to the system based on only the feature stream from original speech.

During online speaker diarization, a situation may occur where a single speaker is being represented by several different models. Such situation leads to worsened diarization results, because the diarization system considers every change of a model to be a change of speakers. In the article we describe a method for detecting this situation and propose several ways of solving it. Experiments show that the most suitable option is treating multiple GMMs as belonging to a single speaker, i.e. updating all of them with the same data every time one of them is assigned a new segment. In that case, there was a relative improvement in Diarization Error Rate of 30.69% in comparison with the baseline system.

In speaker diarization system, it's common to use bottom-up clustering method where the input data is first split in small pieces and then merged the most similar segments until reaching a stopping point. However, it's not easy to ensure that every selection merges the right pair, and these errors tend to deteriorate the post-merging results. In this paper, a fast cluster purification algorithm is introduced after the size of clusters reach to a pre-estimated number K, which is no less than the real number of speakers involved in the conversation, and try to remedy the errors by removing the inappropriate segments into the right cluster. An effective way to estimate the number of speakers is also introduced before the clustering stage. The experiment results show improvement in both the purity of cluster and the diarization error rate (DER) after using the purification algorithm. The purity improves 0.8% and DER reduces 1.14% in average.

The recently proposed speaker diarization technique based on binary keys provides a very fast alternative to state-of-the-art systems with little increase of Diarization Error Rate DER. Although the approach shows great potential, it also presents issues, mainly in the stopping criterion. Therefore, exploring alternative clustering/stopping criterion approaches is needed. Recently some works have addressed the speaker clustering as a global optimization problem in order to tackle the intrinsic issues of the Agglomerative Hierarchical Clustering AHC mainly the local-maximum-based decision making. This paper aims at adapting and applying this new framework to the binary key diarization system. In addition, an analysis of cluster purity across the AHC iterations is done using reference speaker ground-truth labels to select the purer clustering as input for the global framework. Experiments on the REPERE phase 1 test database show improvements of around 6% absolute DER compared to the baseline system output.

In speaker diarization, the most commonly used speaker feature is MFCC, which is also most commonly used speech feature in speech recognition. The newly proposed Power Normalized Cepstrum Coefficients (PNCC) achieves impressive improvement in noisy speech recognition compare to MFCC. It consequently expects a proof for speaker diarization use. In this paper, PNCC is evaluated against MFCC in a meeting domain speaker diarization system. The Diarization Error Rate (DER) shows no positive results with PNCC. This is possibly because of their inhibition in high frequency spectrum which is believed to represents the characteristics of human's voice. An initial model training material select strategy is proposed and used in the speaker diarization system in this work.