Deep biasing for the Transducer can improve the recognition performance of rare words or contextual entities, which is essential in practical applications, especially for streaming Automatic Speech Recognition (ASR). However, deep biasing with large-scale rare words remains challenging, as the performance drops significantly when more distractors exist and there are words with similar grapheme sequences in the bias list. In this paper, we combine the phoneme and textual information of rare words in Transducers to distinguish words with similar pronunciation or spelling. Moreover, the introduction of training with text-only data containing more rare words benefits large-scale deep biasing. The experiments on the LibriSpeech corpus demonstrate that the proposed method achieves state-of-the-art performance on rare word error rate for different scales and levels of bias lists.
Domain adaptation using text-only corpus is challenging in end-to-end(E2E) speech recognition. Adaptation by synthesizing audio from text through TTS is resource-consuming. We present a method to learn Unified Speech-Text Representation in Conformer Transducer(USTR-CT) to enable fast domain adaptation using the text-only corpus. Different from the previous textogram method, an extra text encoder is introduced in our work to learn text representation and is removed during inference, so there is no modification for online deployment. To improve the efficiency of adaptation, single-step and multi-step adaptations are also explored. The experiments on adapting LibriSpeech to SPGISpeech show the proposed method reduces the word error rate(WER) by relatively 44% on the target domain, which is better than those of TTS method and textogram method. Also, it is shown the proposed method can be combined with internal language model estimation(ILME) to further improve the performance.
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 and has high latency. Although it is common to address these issues 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. 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.
Pansharpening is a fundamental issue in remote sensing field. This paper proposes a side information partially guided convolutional sparse coding (SCSC) model for pansharpening. The key idea is to split the low resolution multispectral image into a panchromatic image related feature map and a panchromatic image irrelated feature map, where the former one is regularized by the side information from panchromatic images. With the principle of algorithm unrolling techniques, the proposed model is generalized as a deep neural network, called as SCSC pansharpening neural network (SCSC-PNN). Compared with 13 classic and state-of-the-art methods on three satellites, the numerical experiments show that SCSC-PNN is superior to others. The codes are available at https://github.com/xsxjtu/SCSC-PNN.
Pansharpening is a widely used image enhancement technique for remote sensing. Its principle is to fuse the input high-resolution single-channel panchromatic (PAN) image and low-resolution multi-spectral image and to obtain a high-resolution multi-spectral (HRMS) image. The existing deep learning pansharpening method has two shortcomings. First, features of two input images need to be concatenated along the channel dimension to reconstruct the HRMS image, which makes the importance of PAN images not prominent, and also leads to high computational cost. Second, the implicit information of features is difficult to extract through the manually designed loss function. To this end, we propose a generative adversarial network via the fast guided filter (FGF) for pansharpening. In generator, traditional channel concatenation is replaced by FGF to better retain the spatial information while reducing the number of parameters. Meanwhile, the fusion objects can be highlighted by the spatial attention module. In addition, the latent information of features can be preserved effectively through adversarial training. Numerous experiments illustrate that our network generates high-quality HRMS images that can surpass existing methods, and with fewer parameters.
Utterance-level permutation invariant training (uPIT) has achieved promising progress on single-channel multi-talker speech separation task. Long short-term memory (LSTM) and bidirectional LSTM (BLSTM) are widely used as the separation networks of uPIT, i.e. uPIT-LSTM and uPIT-BLSTM. uPIT-LSTM has lower latency but worse performance, while uPIT-BLSTM has better performance but higher latency. In this paper, we propose using latency-controlled BLSTM (LC-BLSTM) during inference to fulfill low-latency and good-performance speech separation. To find a better training strategy for BLSTM-based separation network, chunk-level PIT (cPIT) and uPIT are compared. The experimental results show that uPIT outperforms cPIT when LC-BLSTM is used during inference. It is also found that the inter-chunk speaker tracing (ST) can further improve the separation performance of uPIT-LC-BLSTM. Evaluated on the WSJ0 two-talker mixed-speech separation task, the absolute gap of signal-to-distortion ratio (SDR) between uPIT-BLSTM and uPIT-LC-BLSTM is reduced to within 0.7 dB.
Convolutional neural network (CNN), in particular the Unet, is a powerful method for medical image segmentation. To date Unet has demonstrated state-of-art performance in many complex medical image segmentation tasks, especially under the condition when the training and testing data share the same distribution (i.e. come from the same source domain). However, in clinical practice, medical images are acquired from different vendors and centers. The performance of a U-Net trained from a particular source domain, when transferred to a different target domain (e.g. different vendor, acquisition parameter), can drop unexpectedly. Collecting a large amount of annotation from each new domain to retrain the U-Net is expensive, tedious, and practically impossible. In this work, we proposed a generic framework to address this problem, consisting of (1) an unpaired generative adversarial network (GAN) for vendor-adaptation, and (2) a Unet for object segmentation. In the proposed Unet-GAN architecture, GAN learns from Unet at the feature level that is segmentation-specific. We used cardiac cine MRI as the example, with three major vendors (Philips, Siemens, and GE) as three domains, while the methodology can be extended to medical images segmentation in general. The proposed method showed significant improvement of the segmentation results across vendors. The proposed Unet-GAN provides an annotation-free solution to the cross-vendor medical image segmentation problem, potentially extending a trained deep learning model to multi-center and multi-vendor use in real clinical scenario.
Long Short-Term Memory (LSTM) is the primary recurrent neural networks architecture for acoustic modeling in automatic speech recognition systems. Residual learning is an efficient method to help neural networks converge easier and faster. In this paper, we propose several types of residual LSTM methods for our acoustic modeling. Our experiments indicate that, compared with classic LSTM, our architecture shows more than 8% relative reduction in Phone Error Rate (PER) on TIMIT tasks. At the same time, our residual fast LSTM approach shows 4% relative reduction in PER on the same task. Besides, we find that all this architecture could have good results on THCHS-30, Librispeech and Switchboard corpora.