We propose a modified teacher-student training for the extraction of frame-wise speaker embeddings that allows for an effective diarization of meeting scenarios containing partially overlapping speech. To this end, a geodesic distance loss is used that enforces the embeddings computed from regions with two active speakers to lie on the shortest path on a sphere between the points given by the d-vectors of each of the active speakers. Using those frame-wise speaker embeddings in clustering-based diarization outperforms segment-level clustering-based diarization systems such as VBx and Spectral Clustering. By extending our approach to a mixture-model-based diarization, the performance can be further improved, approaching the diarization error rates of diarization systems that use a dedicated overlap detection, and outperforming these systems when also employing an additional overlap detection.
We propose a diarization system, that estimates "who spoke when" based on spatial information, to be used as a front-end of a meeting transcription system running on the signals gathered from an acoustic sensor network (ASN). Although the spatial distribution of the microphones is advantageous, exploiting the spatial diversity for diarization and signal enhancement is challenging, because the microphones' positions are typically unknown, and the recorded signals are initially unsynchronized in general. Here, we approach these issues by first blindly synchronizing the signals and then estimating time differences of arrival (TDOAs). The TDOA information is exploited to estimate the speakers' activity, even in the presence of multiple speakers being simultaneously active. This speaker activity information serves as a guide for a spatial mixture model, on which basis the individual speaker's signals are extracted via beamforming. Finally, the extracted signals are forwarded to a speech recognizer. Additionally, a novel initialization scheme for spatial mixture models based on the TDOA estimates is proposed. Experiments conducted on real recordings from the LibriWASN data set have shown that our proposed system is advantageous compared to a system using a spatial mixture model, which does not make use of external diarization information.
Unsupervised speech disentanglement aims at separating fast varying from slowly varying components of a speech signal. In this contribution, we take a closer look at the embedding vector representing the slowly varying signal components, commonly named the speaker embedding vector. We ask, which properties of a speaker's voice are captured and investigate to which extent do individual embedding vector components sign responsible for them, using the concept of Shapley values. Our findings show that certain speaker-specific acoustic-phonetic properties can be fairly well predicted from the speaker embedding, while the investigated more abstract voice quality features cannot.
We propose a modular pipeline for the single-channel separation, recognition, and diarization of meeting-style recordings and evaluate it on the Libri-CSS dataset. Using a Continuous Speech Separation (CSS) system with a TF-GridNet separation architecture, followed by a speaker-agnostic speech recognizer, we achieve state-of-the-art recognition performance in terms of Optimal Reference Combination Word Error Rate (ORC WER). Then, a d-vector-based diarization module is employed to extract speaker embeddings from the enhanced signals and to assign the CSS outputs to the correct speaker. Here, we propose a syntactically informed diarization using sentence- and word-level boundaries of the ASR module to support speaker turn detection. This results in a state-of-the-art Concatenated minimum-Permutation Word Error Rate (cpWER) for the full meeting recognition pipeline.
Many real-life applications of automatic speech recognition (ASR) require processing of overlapped speech. A commonmethod involves first separating the speech into overlap-free streams and then performing ASR on the resulting signals. Recently, the inclusion of a mixture encoder in the ASR model has been proposed. This mixture encoder leverages the original overlapped speech to mitigate the effect of artifacts introduced by the speech separation. Previously, however, the method only addressed two-speaker scenarios. In this work, we extend this approach to more natural meeting contexts featuring an arbitrary number of speakers and dynamic overlaps. We evaluate the performance using different speech separators, including the powerful TF-GridNet model. Our experiments show state-of-the-art performance on the LibriCSS dataset and highlight the advantages of the mixture encoder. Furthermore, they demonstrate the strong separation of TF-GridNet which largely closes the gap between previous methods and oracle separation.
We present LibriWASN, a data set whose design follows closely the LibriCSS meeting recognition data set, with the marked difference that the data is recorded with devices that are randomly positioned on a meeting table and whose sampling clocks are not synchronized. Nine different devices, five smartphones with a single recording channel and four microphone arrays, are used to record a total of 29 channels. Other than that, the data set follows closely the LibriCSS design: the same LibriSpeech sentences are played back from eight loudspeakers arranged around a meeting table and the data is organized in subsets with different percentages of speech overlap. LibriWASN is meant as a test set for clock synchronization algorithms, meeting separation, diarization and transcription systems on ad-hoc wireless acoustic sensor networks. Due to its similarity to LibriCSS, meeting transcription systems developed for the former can readily be tested on LibriWASN. The data set is recorded in two different rooms and is complemented with ground-truth diarization information of who speaks when.
Disentanglement is the task of learning representations that identify and separate factors that explain the variation observed in data. Disentangled representations are useful to increase the generalizability, explainability, and fairness of data-driven models. Only little is known about how well such disentanglement works for speech representations. A major challenge when tackling disentanglement for speech representations are the unknown generative factors underlying the speech signal. In this work, we investigate to what degree speech representations encoding speaker identity can be disentangled. To quantify disentanglement, we identify acoustic features that are highly speaker-variant and can serve as proxies for the factors of variation underlying speech. We find that disentanglement of the speaker embedding is limited when trained with standard objectives promoting disentanglement but can be improved over vanilla representation learning to some extent.
MeetEval is an open-source toolkit to evaluate all kinds of meeting transcription systems. It provides a unified interface for the computation of commonly used Word Error Rates (WERs), specifically cpWER, ORC WER and MIMO WER along other WER definitions. We extend the cpWER computation by a temporal constraint to ensure that only words are identified as correct when the temporal alignment is plausible. This leads to a better quality of the matching of the hypothesis string to the reference string that more closely resembles the actual transcription quality, and a system is penalized if it provides poor time annotations. Since word-level timing information is often not available, we present a way to approximate exact word-level timings from segment-level timings (e.g., a sentence) and show that the approximation leads to a similar WER as a matching with exact word-level annotations. At the same time, the time constraint leads to a speedup of the matching algorithm, which outweighs the additional overhead caused by processing the time stamps.
Due to the high variation in the application requirements of sound event detection (SED) systems, it is not sufficient to evaluate systems only in a single operating mode. Therefore, the community recently adopted the polyphonic sound detection score (PSDS) as an evaluation metric, which is the normalized area under the PSD receiver operating characteristic (PSD-ROC). It summarizes the system performance over a range of operating modes resulting from varying the decision threshold that is used to translate the system output scores into a binary detection output. Hence, it provides a more complete picture of the overall system behavior and is less biased by specific threshold tuning. However, besides the decision threshold there is also the post-processing that can be changed to enter another operating mode. In this paper we propose the post-processing independent PSDS (piPSDS) as a generalization of the PSDS. Here, the post-processing independent PSD-ROC includes operating points from varying post-processings with varying decision thresholds. Thus, it summarizes even more operating modes of an SED system and allows for system comparison without the need of implementing a post-processing and without a bias due to different post-processings. While piPSDS can in principle combine different types of post-processing, we hear, as a first step, present median filter independent PSDS (miPSDS) results for this year's DCASE Challenge Task4a systems. Source code is publicly available in our sed_scores_eval package (https://github.com/fgnt/sed_scores_eval).
Multi-speaker automatic speech recognition (ASR) is crucial for many real-world applications, but it requires dedicated modeling techniques. Existing approaches can be divided into modular and end-to-end methods. Modular approaches separate speakers and recognize each of them with a single-speaker ASR system. End-to-end models process overlapped speech directly in a single, powerful neural network. This work proposes a middle-ground approach that leverages explicit speech separation similarly to the modular approach but also incorporates mixture speech information directly into the ASR module in order to mitigate the propagation of errors made by the speech separator. We also explore a way to exchange cross-speaker context information through a layer that combines information of the individual speakers. Our system is optimized through separate and joint training stages and achieves a relative improvement of 7% in word error rate over a purely modular setup on the SMS-WSJ task.