This paper focuses on multi-enrollment speaker recognition which naturally occurs in the task of online speaker clustering, and studies the properties of different scoring back-ends in this scenario. First, we show that popular cosine scoring suffers from poor score calibration with a varying number of enrollment utterances. Second, we propose a simple replacement for cosine scoring based on an extremely constrained version of probabilistic linear discriminant analysis (PLDA). The proposed model improves over the cosine scoring for multi-enrollment recognition while keeping the same performance in the case of one-to-one comparisons. Finally, we consider an online speaker clustering task where each step naturally involves multi-enrollment recognition. We propose an online clustering algorithm allowing us to take benefits from the PLDA model such as the ability to handle uncertainty and better score calibration. Our experiments demonstrate the effectiveness of the proposed algorithm.
The pre-trained multi-lingual XLSR model generalizes well for language identification after fine-tuning on unseen languages. However, the performance significantly degrades when the languages are not very distinct from each other, for example, in the case of dialects. Low resource dialect classification remains a challenging problem to solve. We present a new data augmentation method that leverages model training dynamics of individual data points to improve sampling for latent mixup. The method works well in low-resource settings where generalization is paramount. Our datamaps-based mixup technique, which we call Map-Mix improves weighted F1 scores by 2% compared to the random mixup baseline and results in a significantly well-calibrated model. The code for our method is open sourced on https://github.com/skit-ai/Map-Mix.
Audio-visual speech recognition (AVSR) has gained remarkable success for ameliorating the noise-robustness of speech recognition. Mainstream methods focus on fusing audio and visual inputs to obtain modality-invariant representations. However, such representations are prone to over-reliance on audio modality as it is much easier to recognize than video modality in clean conditions. As a result, the AVSR model underestimates the importance of visual stream in face of noise corruption. To this end, we leverage visual modality-specific representations to provide stable complementary information for the AVSR task. Specifically, we propose a reinforcement learning (RL) based framework called MSRL, where the agent dynamically harmonizes modality-invariant and modality-specific representations in the auto-regressive decoding process. We customize a reward function directly related to task-specific metrics (i.e., word error rate), which encourages the MSRL to effectively explore the optimal integration strategy. Experimental results on the LRS3 dataset show that the proposed method achieves state-of-the-art in both clean and various noisy conditions. Furthermore, we demonstrate the better generality of MSRL system than other baselines when test set contains unseen noises.
This paper summarizes the outcomes from the ISCSLP 2022 Intelligent Cockpit Speech Recognition Challenge (ICSRC). We first address the necessity of the challenge and then introduce the associated dataset collected from a new-energy vehicle (NEV) covering a variety of cockpit acoustic conditions and linguistic contents. We then describe the track arrangement and the baseline system. Specifically, we set up two tracks in terms of allowed model/system size to investigate resource-constrained and -unconstrained setups, targeting to vehicle embedded as well as cloud ASR systems respectively. Finally we summarize the challenge results and provide the major observations from the submitted systems.
To let the state-of-the-art end-to-end ASR model enjoy data efficiency, as well as much more unpaired text data by multi-modal training, one needs to address two problems: 1) the synchronicity of feature sampling rates between speech and language (aka text data); 2) the homogeneity of the learned representations from two encoders. In this paper we propose to employ a novel bidirectional attention mechanism (BiAM) to jointly learn both ASR encoder (bottom layers) and text encoder with a multi-modal learning method. The BiAM is to facilitate feature sampling rate exchange, realizing the quality of the transformed features for the one kind to be measured in another space, with diversified objective functions. As a result, the speech representations are enriched with more linguistic information, while the representations generated by the text encoder are more similar to corresponding speech ones, and therefore the shared ASR models are more amenable for unpaired text data pretraining. To validate the efficacy of the proposed method, we perform two categories of experiments with or without extra unpaired text data. Experimental results on Librispeech corpus show it can achieve up to 6.15% word error rate reduction (WERR) with only paired data learning, while 9.23% WERR when more unpaired text data is employed.
Noise robustness in keyword spotting remains a challenge as many models fail to overcome the heavy influence of noises, causing the deterioration of the quality of feature embeddings. We proposed a contrastive regularization method called Inter-Intra Contrastive Regularization (I2CR) to improve the feature representations by guiding the model to learn the fundamental speech information specific to the cluster. This involves maximizing the similarity across Intra and Inter samples of the same class. As a result, it pulls the instances closer to more generalized representations that form more prominent clusters and reduces the adverse impact of noises. We show that our method provides consistent improvements in accuracy over different backbone model architectures under different noise environments. We also demonstrate that our proposed framework has improved the accuracy of unseen out-of-domain noises and unseen variant noise SNRs. This indicates the significance of our work with the overall refinement in noise robustness.
Nuclear Magnetic Resonance (NMR) is used in structural biology to experimentally determine the structure of proteins, which is used in many areas of biology and is an important part of drug development. Unfortunately, NMR data can cost thousands of dollars per sample to collect and it can take a specialist weeks to assign the observed resonances to specific chemical groups. There has thus been growing interest in the NMR community to use deep learning to automate NMR data annotation. Due to similarities between NMR and audio data, we propose that methods used in acoustic signal processing can be applied to NMR as well. Using a simulated amino acid dataset, we show that by swapping out filter banks with a trainable convolutional encoder, acoustic signal embeddings from speaker verification models can be used for amino acid classification in 2D NMR spectra by treating each amino acid as a unique speaker. On an NMR dataset comparable in size with of 46 hours of audio, we achieve a classification performance of 97.7% on a 20-class problem. We also achieve a 23% relative improvement by using an acoustic embedding model compared to an existing NMR-based model.
Continuously learning new classes without catastrophic forgetting is a challenging problem for on-device environmental sound classification given the restrictions on computation resources (e.g., model size, running memory). To address this issue, we propose a simple and efficient continual learning method. Our method selects the historical data for the training by measuring the per-sample classification uncertainty. Specifically, we measure the uncertainty by observing how the classification probability of data fluctuates against the parallel perturbations added to the classifier embedding. In this way, the computation cost can be significantly reduced compared with adding perturbation to the raw data. Experimental results on the DCASE 2019 Task 1 and ESC-50 dataset show that our proposed method outperforms baseline continual learning methods on classification accuracy and computational efficiency, indicating our method can efficiently and incrementally learn new classes without the catastrophic forgetting problem for on-device environmental sound classification.
Intermediate layer output (ILO) regularization by means of multitask training on encoder side has been shown to be an effective approach to yielding improved results on a wide range of end-to-end ASR frameworks. In this paper, we propose a novel method to do ILO regularized training differently. Instead of using conventional multitask methods that entail more training overhead, we directly make the intermediate layer output as input to the decoder, that is, our decoder not only accepts the output of the final encoder layer as input, it also takes the output of the encoder ILO as input during training. With the proposed method, as both encoder and decoder are simultaneously "regularized", the network is more sufficiently trained, consistently leading to improved results, over the ILO-based CTC method, as well as over the original attention-based modeling method without the proposed method employed.
Internal Language Model Estimation (ILME) based language model (LM) fusion has been shown significantly improved recognition results over conventional shallow fusion in both intra-domain and cross-domain speech recognition tasks. In this paper, we attempt to apply our ILME method to cross-domain code-switching speech recognition (CSSR) work. Specifically, our curiosity comes from several aspects. First, we are curious about how effective the ILME-based LM fusion is for both intra-domain and cross-domain CSSR tasks. We verify this with or without merging two code-switching domains. More importantly, we train an end-to-end (E2E) speech recognition model by means of merging two monolingual data sets and observe the efficacy of the proposed ILME-based LM fusion for CSSR. Experimental results on SEAME that is from Southeast Asian and another Chinese Mainland CS data set demonstrate the effectiveness of the proposed ILME-based LM fusion method.