Hyung Yong Kim

Knowledge distillation (KD), best known as an effective method for model compression, aims at transferring the knowledge of a bigger network (teacher) to a much smaller network (student). Conventional KD methods usually employ the teacher model trained in a supervised manner, where output labels are treated only as targets. Extending this supervised scheme further, we introduce a new type of teacher model for connectionist temporal classification (CTC)-based sequence models, namely Oracle Teacher, that leverages both the source inputs and the output labels as the teacher model's input. Since the Oracle Teacher learns a more accurate CTC alignment by referring to the target information, it can provide the student with more optimal guidance. One potential risk for the proposed approach is a trivial solution that the model's output directly copies the target input. Based on a many-to-one mapping property of the CTC algorithm, we present a training strategy that can effectively prevent the trivial solution and thus enables utilizing both source and target inputs for model training. Extensive experiments are conducted on two sequence learning tasks: speech recognition and scene text recognition. From the experimental results, we empirically show that the proposed model improves the students across these tasks while achieving a considerable speed-up in the teacher model's training time.

Pre-training of self-supervised learning (SSL) generally shows a good performance on various speech processing tasks. However, this pre-training scheme may lead to a sub-optimal solution for fine-tuning a specific task, such as automatic speech recognition (ASR). In order to provide a more optimal pre-trained model for ASR, we introduce an ASR-Specific hidden-unit BERT with self-training, namely ASBERT. Motivated by self-training, we extract linguistic-related pseudo labels from the fine-tuned model, and these labels are used in the next pre-training procedure. Experimental results on LibriSpeech test-clean and test-other datasets show that ASBERT without language model (LM) outperforms the conventional SSL and self-training model, achieving a 6.3/2.0% and 15.4/13.2% relatively word error rate reduction (RERR). Moreover, without using pseudo-transcription, ASBERT yields comparable performance to the conventional self-training method.

Recently, a conventional neural decoder for speech codec has shown promising performance. However, it typically requires some prior knowledge of decoding such as bit allocation or dequantization scheme, which is not a universal solution for many different kinds of speech codecs. In order to address this limitation, we propose a neurally optimized decoder based on a generative model which can directly reconstruct the speech from the bitstream without a prior knowledge. The proposed decoder mainly consists of two components: 1) a dequantization model to group and dequantize related bits from the bitstream and 2) a generative model to restore the speech conditioned on the output of the dequantization model. Through experiments with mixed excitation linear prediction (MELP), Advanced multi-band excitation (AMBE), and SPEEX at around 2.4 kb/s, it is showed that the proposed model showed better performance in most of the objective and subjective evaluation compared to the conventional speech codecs.

In recent years, there has been a great deal of research in developing end-to-end speech recognition models, which enable simplifying the traditional pipeline and achieving promising results. Despite their remarkable performance improvements, end-to-end models typically require expensive computational cost to show successful performance. To reduce this computational burden, knowledge distillation (KD), which is a popular model compression method, has been used to transfer knowledge from a deep and complex model (teacher) to a shallower and simpler model (student). Previous KD approaches have commonly designed the architecture of the student by reducing the width per layer or the number of layers of the teacher. This structural reduction scheme might limit the flexibility of model selection since the student model structure should be similar to that of the given teacher. To cope with this limitation, we propose a KD method for end-to-end speech recognition, namely TutorNet, that applies KD techniques across different types of neural networks at the hidden representation-level as well as the output-level. For concrete realizations, we firstly apply representation-level knowledge distillation (RKD) during the initialization step, and then apply the softmax-level knowledge distillation (SKD) combined with the original task learning. When the student is trained with RKD, we make use of frame weighting that points out the frames to which the teacher pays more attention. Through a number of experiments, it is verified that TutorNet not only distills the knowledge between networks with different topologies but also significantly contributes to improving the performance of the distilled student.

Recently, generative adversarial networks (GANs) have been successfully applied to speech enhancement. However, there still remain two issues that need to be addressed: (1) GAN-based training is typically unstable due to its non-convex property, and (2) most of the conventional methods do not fully take advantage of the speech characteristics, which could result in a sub-optimal solution. In order to deal with these problems, we propose a progressive generator that can handle the speech in a multi-resolution fashion. Additionally, we propose a multi-scale discriminator that discriminates the real and generated speech at various sampling rates to stabilize GAN training. The proposed structure was compared with the conventional GAN-based speech enhancement algorithms using the VoiceBank-DEMAND dataset. Experimental results showed that the proposed approach can make the training faster and more stable, which improves the performance on various metrics for speech enhancement.

For multi-channel speech recognition, speech enhancement techniques such as denoising or dereverberation are conventionally applied as a front-end processor. Deep learning-based front-ends using such techniques require aligned clean and noisy speech pairs which are generally obtained via data simulation. Recently, several joint optimization techniques have been proposed to train the front-end without parallel data within an end-to-end automatic speech recognition (ASR) scheme. However, the ASR objective is sub-optimal and insufficient for fully training the front-end, which still leaves room for improvement. In this paper, we propose a novel approach which incorporates flow-based density estimation for the robust front-end using non-parallel clean and noisy speech. Experimental results on the CHiME-4 dataset show that the proposed method outperforms the conventional techniques where the front-end is trained only with ASR objective.