In speaker recognition, where speech segments are mapped to embeddings on the unit hypersphere, two scoring backends are commonly used, namely cosine scoring or PLDA. Both have advantages and disadvantages, depending on the context. Cosine scoring follows naturally from the spherical geometry, but for PLDA the blessing is mixed -- length normalization Gaussianizes the between-speaker distribution, but violates the assumption of a speaker-independent within-speaker distribution. We propose PSDA, an analogue to PLDA that uses Von Mises-Fisher distributions on the hypersphere for both within and between-class distributions. We show how the self-conjugacy of this distribution gives closed-form likelihood-ratio scores, making it a drop-in replacement for PLDA at scoring time. All kinds of trials can be scored, including single-enroll and multi-enroll verification, as well as more complex likelihood-ratios that could be used in clustering and diarization. Learning is done via an EM-algorithm with closed-form updates. We explain the model and present some first experiments.
This paper addresses the task of performing robust and reliable time-delay estimation in audio-signals in noisy and reverberating environments. In contrast to the popular signal processing based methods, this paper proposes machine learning based method, i.e., a semi-causal convolutional neural network consisting of a set of causal and anti-causal layers with a novel correlation-based objective function. The causality in the network ensures non-leakage of representations from future time-intervals and the proposed loss function makes the network generate sequences with high correlation at the actual time delay. The proposed approach is also intrinsically interpretable as it does not lose time information. Even a shallow convolution network is able to capture local patterns in sequences, while also correlating them globally. SyncNet outperforms other classical approaches in estimating mutual time delays for different types of audio signals including pulse, speech and musical beats.
This paper evaluates a wide range of audio-based deep learning frameworks applied to the breathing, cough, and speech sounds for detecting COVID-19. In general, the audio recording inputs are transformed into low-level spectrogram features, then they are fed into pre-trained deep learning models to extract high-level embedding features. Next, the dimension of these high-level embedding features are reduced before finetuning using Light Gradient Boosting Machine (LightGBM) as a back-end classification. Our experiments on the Second DiCOVA Challenge achieved the highest Area Under the Curve (AUC), F1 score, sensitivity score, and specificity score of 89.03%, 64.41%, 63.33%, and 95.13%, respectively. Based on these scores, our method outperforms the state-of-the-art systems, and improves the challenge baseline by 4.33%, 6.00% and 8.33% in terms of AUC, F1 score and sensitivity score, respectively.
In this paper, we present the speaker diarization system for the Multi-channel Multi-party Meeting Transcription Challenge (M2MeT) from team DKU_DukeECE. As the highly overlapped speech exists in the dataset, we employ an x-vector-based target-speaker voice activity detection (TS-VAD) to find the overlap between speakers. For the single-channel scenario, we separately train a model for each of the 8 channels and fuse the results. We also employ the cross-channel self-attention to further improve the performance, where the non-linear spatial correlations between different channels are learned and fused. Experimental results on the evaluation set show that the single-channel TS-VAD reduces the DER by over 75% from 12.68\% to 3.14%. The multi-channel TS-VAD further reduces the DER by 28% and achieves a DER of 2.26%. Our final submitted system achieves a DER of 2.98% on the AliMeeting test set, which ranks 1st in the M2MET challenge.
Warning: this paper contains material which may be offensive or upsetting. While much of recent work has focused on the detection of hate speech and overtly offensive content, very little research has explored the more subtle but equally harmful language in the form of implied stereotypes. This is a challenging domain, made even more so by the fact that humans often struggle to understand and reason about stereotypes. We build on existing literature and present CO-STAR (COnceptualisation of STereotypes for Analysis and Reasoning), a novel framework which encodes the underlying concepts of implied stereotypes. We also introduce the CO-STAR training data set, which contains just over 12K structured annotations of implied stereotypes and stereotype conceptualisations, and achieve state-of-the-art results after training and manual evaluation. The CO-STAR models are, however, limited in their ability to understand more complex and subtly worded stereotypes, and our research motivates future work in developing models with more sophisticated methods for encoding common-sense knowledge.
Language models trained on large-scale unfiltered datasets curated from the open web acquire systemic biases, prejudices, and harmful views from their training data. We present a methodology for programmatically identifying and removing harmful text from web-scale datasets. A pretrained language model is used to calculate the log-likelihood of researcher-written trigger phrases conditioned on a specific document, which is used to identify and filter documents from the dataset. We demonstrate that models trained on this filtered dataset exhibit lower propensity to generate harmful text, with a marginal decrease in performance on standard language modeling benchmarks compared to unfiltered baselines. We provide a partial explanation for this performance gap by surfacing examples of hate speech and other undesirable content from standard language modeling benchmarks. Finally, we discuss the generalization of this method and how trigger phrases which reflect specific values can be used by researchers to build language models which are more closely aligned with their values.
Timbre representations of musical instruments, essential for diverse applications such as musical audio synthesis and separation, might be learned as bottleneck features from an instrumental recognition model. Given the similarities between speaker recognition and musical instrument recognition, in this paper, we investigate how to adapt successful speaker recognition algorithms to musical instrument recognition to learn meaningful instrumental timbre representations. To address the mismatch between musical audio and models devised for speech, we introduce a group of trainable filters to generate proper acoustic features from input raw waveforms, making it easier for a model to be optimized in an input-agnostic and end-to-end manner. Through experiments on both the NSynth and RWC databases in both musical instrument closed-set identification and open-set verification scenarios, the modified speaker recognition model was capable of generating discriminative embeddings for instrument and instrument-family identities. We further conducted extensive experiments to characterize the encoded information in learned timbre embeddings.
Deep neural networks provide effective solutions to small-footprint keyword spotting (KWS). However, if training data is limited, it remains challenging to achieve robust and highly accurate KWS in real-world scenarios where unseen sounds that are out of the training data are frequently encountered. Most conventional methods aim to maximize the classification accuracy on the training set, without taking the unseen sounds into account. To enhance the robustness of the deep neural networks based KWS, in this paper, we introduce a new loss function, named the maximization of the area under the receiver-operating-characteristic curve (AUC). The proposed method not only maximizes the classification accuracy of keywords on the closed training set, but also maximizes the AUC score for optimizing the performance of non-keyword segments detection. Experimental results on the Google Speech Commands dataset v1 and v2 show that our method achieves new state-of-the-art performance in terms of most evaluation metrics.
Add to Chrome
Add to Firefox