Search

This paper addresses the persistent challenge in Keyword Spotting (KWS), a fundamental component in speech technology, regarding the acquisition of substantial labeled data for training. Given the difficulty in obtaining large quantities of positive samples and the laborious process of collecting new target samples when the keyword changes, we introduce a novel approach combining unsupervised contrastive learning and a unique augmentation-based technique. Our method allows the neural network to train on unlabeled data sets, potentially improving performance in downstream tasks with limited labeled data sets. We also propose that similar high-level feature representations should be employed for speech utterances with the same keyword despite variations in speed or volume. To achieve this, we present a speech augmentation-based unsupervised learning method that utilizes the similarity between the bottleneck layer feature and the audio reconstructing information for auxiliary training. Furthermore, we propose a compressed convolutional architecture to address potential redundancy and non-informative information in KWS tasks, enabling the model to simultaneously learn local features and focus on long-term information. This method achieves strong performance on the Google Speech Commands V2 Dataset. Inspired by recent advancements in sign spotting and spoken term detection, our method underlines the potential of our contrastive learning approach in KWS and the advantages of Query-by-Example Spoken Term Detection strategies. The presented CAB-KWS provide new perspectives in the field of KWS, demonstrating effective ways to reduce data collection efforts and increase the system's robustness., Comment: This paper has been accepted by the ICPR2024

Efficiently handling Automatic Identification System (AIS) data is vital for enhancing maritime safety and navigation, yet is hindered by the system's high volume and error-prone datasets. This paper introduces the Automatic Identification System Database (AISdb), a novel tool designed to address the challenges of processing and analyzing AIS data. AISdb is a comprehensive, open-source platform that enables the integration of AIS data with environmental datasets, thus enriching analyses of vessel movements and their environmental impacts. By facilitating AIS data collection, cleaning, and spatio-temporal querying, AISdb significantly advances AIS data research. Utilizing AIS data from various sources, AISdb demonstrates improved handling and analysis of vessel information, contributing to enhancing maritime safety, security, and environmental sustainability efforts.

Efficient sterilization of pathogens with cleaner methods is a critical concern for environmental disinfection and clinical anti-infective treatment. Plasma-activated water (PAW) is a promising alternative to chemical disinfectants and antibiotics for its strong sterilization ability and not inducing any acute toxicity, and only water and air are consumed during production. For more efficient water activation, plasma sources are commonly placed near or fully in contact with water as possible, but the risks of electrode corrosion and metal contamination of water threaten the safety and stability of PAW production. Herein, plasma-activated gas rich in high-valence NOx is generated by a hybrid plasma configuration and introduced into water for off-site PAW production. Plasma-generated O3 is found to dominate the gas-phase reactions for the formation of high-valence NOx. With the time-evolution of O3 concentration, gaseous NO3 radicals are produced behind N2O5 formation, but will be decomposed before N2O5 quenching. By decoupling the roles of gaseous NO3, N2O5, and O3 in the water activation, results show that short-lived aqueous species induced by gaseous NO3 radicals play the most crucial role in PAW sterilization, and the acidic environment induced by N2O5 is also essential. Moreover, SEM photographs and biomacromolecule leakage assays demonstrate that PAW disrupts the cell membranes of bacteria to achieve inactivation. In real-life applications, an integrated device for off-site PAW production with a yield of 2 L/h and a bactericidal efficiency of >99.9% is developed. The PAW of 50mL produced in 3 minutes using this device is more effective in disinfection than 0.5% NaClO and 3% H2O2 with the same bacterial contact time. This work provides new avenues for efficient PAW production and deepens insights into the fundamental processes that govern the reactive chemistry in PAW sterilization.

Evaluations of model editing currently only use the `next few token' completions after a prompt. As a result, the impact of these methods on longer natural language generation is largely unknown. We introduce long-form evaluation of model editing (LEME) a novel evaluation protocol that measures the efficacy and impact of model editing in long-form generative settings. Our protocol consists of a machine-rated survey and a classifier which correlates well with human ratings. Importantly, we find that our protocol has very little relationship with previous short-form metrics (despite being designed to extend efficacy, generalization, locality, and portability into a long-form setting), indicating that our method introduces a novel set of dimensions for understanding model editing methods. Using this protocol, we benchmark a number of model editing techniques and present several findings including that, while some methods (ROME and MEMIT) perform well in making consistent edits within a limited scope, they suffer much more from factual drift than other methods. Finally, we present a qualitative analysis that illustrates common failure modes in long-form generative settings including internal consistency, lexical cohesion, and locality issues.

This work describes an encoder pre-training procedure using frame-wise label to improve the training of streaming recurrent neural network transducer (RNN-T) model. Streaming RNN-T trained from scratch usually performs worse than non-streaming RNN-T. Although it is common to address this issue through pre-training components of RNN-T with other criteria or frame-wise alignment guidance, the alignment is not easily available in end-to-end manner. In this work, frame-wise alignment, used to pre-train streaming RNN-T's encoder, is generated without using a HMM-based system. Therefore an all-neural framework equipping HMM-free encoder pre-training is constructed. This is achieved by expanding the spikes of CTC model to their left/right blank frames, and two expanding strategies are proposed. To our best knowledge, this is the first work to simulate HMM-based frame-wise label using CTC model for pre-training. Experiments conducted on LibriSpeech and MLS English tasks show the proposed pre-training procedure, compared with random initialization, reduces the WER by relatively 5%~11% and the emission latency by 60 ms. Besides, the method is lexicon-free, so it is friendly to new languages without manually designed lexicon., Comment: Accepted by Interspeech 2021

In this paper, we apply the NetFV and NetVLAD layers for the end-to-end language identification task. NetFV and NetVLAD layers are the differentiable implementations of the standard Fisher Vector and Vector of Locally Aggregated Descriptors (VLAD) methods, respectively. Both of them can encode a sequence of feature vectors into a fixed dimensional vector which is very important to process those variable-length utterances. We first present the relevances and differences between the classical i-vector and the aforementioned encoding schemes. Then, we construct a flexible end-to-end framework including a convolutional neural network (CNN) architecture and an encoding layer (NetFV or NetVLAD) for the language identification task. Experimental results on the NIST LRE 2007 close-set task show that the proposed system achieves significant EER reductions against the conventional i-vector baseline and the CNN temporal average pooling system, respectively., Comment: Accepted for ISCSLP 2018

The classical i-vectors and the latest end-to-end deep speaker embeddings are the two representative categories of utterance-level representations in automatic speaker verification systems. Traditionally, once i-vectors or deep speaker embeddings are extracted, we rely on an extra length normalization step to normalize the representations into unit-length hyperspace before back-end modeling. In this paper, we explore how the neural network learns length-normalized deep speaker embeddings in an end-to-end manner. To this end, we add a length normalization layer followed by a scale layer before the output layer of the common classification network. We conducted experiments on the verification task of the Voxceleb1 dataset. The results show that integrating this simple step in the end-to-end training pipeline significantly boosts the performance of speaker verification. In the testing stage of our L2-normalized end-to-end system, a simple inner-product can achieve the state-of-the-art., Comment: Accepted for Interspeech 2018

In this paper, we explore the encoding/pooling layer and loss function in the end-to-end speaker and language recognition system. First, a unified and interpretable end-to-end system for both speaker and language recognition is developed. It accepts variable-length input and produces an utterance level result. In the end-to-end system, the encoding layer plays a role in aggregating the variable-length input sequence into an utterance level representation. Besides the basic temporal average pooling, we introduce a self-attentive pooling layer and a learnable dictionary encoding layer to get the utterance level representation. In terms of loss function for open-set speaker verification, to get more discriminative speaker embedding, center loss and angular softmax loss is introduced in the end-to-end system. Experimental results on Voxceleb and NIST LRE 07 datasets show that the performance of end-to-end learning system could be significantly improved by the proposed encoding layer and loss function., Comment: Accepted for Speaker Odyssey 2018