Recent sparse multi-view scene reconstruction advances like DUSt3R and MASt3R no longer require camera calibration and camera pose estimation. However, they only process a pair of views at a time to infer pixel-aligned pointmaps. When dealing with more than two views, a combinatorial number of error prone pairwise reconstructions are usually followed by an expensive global optimization, which often fails to rectify the pairwise reconstruction errors. To handle more views, reduce errors, and improve inference time, we propose the fast single-stage feed-forward network MV- DUSt3R. At its core are multi-view decoder blocks which exchange information across any number of views while considering one reference view. To make our method robust to reference view selection, we further propose MV-DUSt3R+, which employs cross-reference-view blocks to fuse information across different reference view choices. To further enable novel view synthesis, we extend both by adding and jointly training Gaussian splatting heads. Experiments on multi-view stereo reconstruction, multi-view pose estimation, and novel view synthesis confirm that our methods improve significantly upon prior art. Code released.1

Deep neural networks for image super-resolution (SR) have demonstrated superior performance. However, the large memory and computation consumption hinders their deployment on resource-constrained devices. Binary neural networks (BNNs), which quantize the floating point weights and activations to 1-bit can significantly reduce the cost. Although BNNs for image classification have made great progress these days, existing BNNs for SR still suffer from a large performance gap between the FP SR networks. To this end, we observe the activation distribution in SR networks and find much larger pixel-to-pixel, channel-to-channel, layer-to-layer, and image-to-image variation in the acti-vation distribution than image classification networks. However, existing BNNs for SR fail to capture these variations that contain rich information for image reconstruction, leading to inferior performance. To address this problem, we propose SCALES, a binarization method for SR networks that consists of the layer-wise scaling factor, the spatial re-scaling method, and the channel-wise re-scaling method, capturing the layer-wise, pixel-wise, and channel-wise variations efficiently in an input-dependent manner. We evaluate our method across different network architectures and datasets. For CNN-based SR networks, our binarization method SCALES outperforms the prior art method by 0.2dB with fewer parameters and operations. With SCALES, we achieve the first accurate binary Transformer-based SR network, improving PSNR by more than 1dB compared to the baseline method.

We present a novel approach to shape editing, building on recent progress in 3D reconstruction from multi-view images. We formulate shape editing as a conditional reconstruction problem, where the model must reconstruct the input shape with the exception of a specified 3D region, in which the geometry should be generated from the conditional signal. To this end, we train a conditional Large Reconstruction Model (LRM) for masked reconstruction, using multi-view consistent masks rendered from a randomly generated 3D occlusion, and using one clean viewpoint as the conditional signal. During inference, we manually define a 3D region to edit and provide an edited image from a canonical viewpoint to fill that region. We demonstrate that, in just a single forward pass, our method not only preserves the input geometry in the unmasked region through reconstruction capabilities on par with SoTA, but is also expressive enough to perform a variety of mesh edits from a single image guidance that past works struggle with, while being 2-10x faster than the top-performing prior work.

Extending the context length of Language Models (LMs) by improving Rotary Position Embedding (RoPE) has become a trend. While prior works mainly address RoPE’s limitations within attention, this paper uncovers the adverse effects on length generalization from nearly all parts of LMs. Using Discrete Signal Processing theory, we show that RoPE enables periodic attention by implicitly achieving Non-Uniform Discrete Fourier Transform. However, this periodicity is undermined by the spectrum damage caused by: 1) linear layers and activation functions outside of attention; 2) insufficiently trained frequency components brought by time-domain truncation. Building on our observations, we propose Fourier Position Embedding (FoPE), which enhances attention’s frequency-domain properties to improve both its periodic extension and length generalization. FoPE constructs Fourier Series and zero-outs the destructive frequency components, increasing model robustness against the spectrum damage. Experiments across various model scales and benchmarks show that, within varying context windows, FoPE maintains a more stable performance compared to other baselines. Several analyses and ablations bring further support to our method and theoretical modeling.

3D scene reconstruction from a single measurement is challenging, especially in the presence of occluded regions and specular materials, such as mirrors. We address these challenges by leveraging single-photon lidars. These lidars estimate depth from light that is emitted into the scene and reflected directly back to the sensor. However, they can also measure light that bounces multiple times in the scene before reaching the sensor. This multi-bounce light contains additional information that can be used to recover dense depth, occluded geometry, and material properties. Prior work with single-photon lidar, however, has only demonstrated these use cases when a laser sequentially illuminates one scene point at a time. We instead focus on the more practical – and challenging – scenario of illuminating multiple scene points simultaneously. The complexity of light transport due to the combined effects of multiplexed illumination, two-bounce light, shadows, and specular reflections is challenging to invert analytically. Instead, we propose a data-driven method to invert light transport in single-photon lidar. To enable this approach, we create the first large-scale simulated dataset of ~100k lidar transients for indoor scenes. We use this dataset to learn a prior on complex light transport, enabling measured two-bounce light to be decomposed into the constituent contributions from each laser spot. Finally, we experimentally demonstrate how this decomposed light can be used to infer 3D geometry in scenes with occlusions and mirrors from a single measurement. Our code and dataset are released on our project webpage.

We present Make-A-Texture, a new framework that efficiently synthesizes high-resolution texture maps from textual prompts for given 3D geometries. Our approach progressively generates textures that are consistent across multiple viewpoints with a depth-aware inpainting diffusion model, in an optimized sequence of viewpoints determined by an automatic view selection algorithm. A significant feature of our method is its remarkable efficiency, achieving a full texture generation within an end-to-end runtime of just 3.07 seconds on a single NVIDIA H100 GPU, significantly outperforming existing methods. Such an acceleration is achieved by optimizations in the diffusion model and a specialized backprojection method. Moreover, our method reduces the artifacts in the backprojection phase, by selectively masking out non-frontal faces, and internal faces of open-surfaced objects. Experimental results demonstrate that Make-A-Texture matches or exceeds the quality of other state-of-the-art methods. Our work significantly improves the applicability and practicality of texture generation models for real-world 3D content creation, including interactive creation and text-guided texture editing.

3D reconstruction from a single-view is challenging because of the ambiguity from monocular cues and lack of information about occluded regions. Neural radiance fields (NeRF), while popular for view synthesis and 3D reconstruction, are typically reliant on multi-view images. Existing methods for single-view 3D reconstruction with NeRF rely on either data priors to hallucinate views of occluded regions, which may not be physically accurate, or shadows observed by RGB cameras, which are difficult to detect in ambient light and low albedo backgrounds. We propose using time-of-flight data captured by a single-photon avalanche diode to overcome these limitations. Our method models two-bounce optical paths with NeRF, using lidar transient data for supervision. By leveraging the advantages of both NeRF and two-bounce light measured by lidar, we demonstrate that we can reconstruct visible and occluded geometry without data priors or reliance on controlled ambient lighting or scene albedo. In addition, we demonstrate improved generalization under practical constraints on sensor spatial- and temporal-resolution. We believe our method is a promising direction as single-photon lidars become ubiquitous on consumer devices, such as phones, tablets, and headsets.

Personalized text-to-image (T2I) synthesis based on diffusion models has attracted significant attention in recent research. However, existing methods primarily concentrate on customizing subjects or styles, neglecting the exploration of global geometry. In this study, we propose an approach that focuses on the customization of 360-degree panoramas, which inherently possess global geometric properties, using a T2I diffusion model. To achieve this, we curate a paired image-text dataset specifically designed for the task and subsequently employ it to fine-tune a pre-trained T2I diffusion model with LoRA. Nevertheless, the fine-tuned model alone does not ensure the continuity between the leftmost and rightmost sides of the synthesized images, a crucial characteristic of 360-degree panoramas. To address this issue, we propose a method called StitchDiffusion. Specifically, we perform pre-denoising operations twice at each time step of the denoising process on the stitch block consisting of the leftmost and rightmost image regions. Furthermore, a global cropping is adopted to synthesize seamless 360-degree panoramas. Experimental results demonstrate the effectiveness of our customized model combined with the proposed StitchDiffusion in generating high-quality 360-degree panoramic images. Moreover, our customized model exhibits exceptional generalization ability in producing scenes unseen in the fine-tuning dataset. Code is available at https://github.com/littlewhitesea/StitchDiffusion.

Video restoration aims to restore high-quality frames from low-quality frames. Different from single image restoration, video restoration generally requires to utilize temporal information from multiple adjacent but usually misaligned video frames. Existing deep methods generally tackle with this by exploiting a sliding window strategy or a recurrent architecture, which are restricted by frame-by-frame restoration. In this paper, we propose a Video Restoration Transformer (VRT) with parallel frame prediction ability. More specifically, VRT is composed of multiple scales, each of which consists of two kinds of modules: temporal reciprocal self attention (TRSA) and parallel warping. TRSA divides the video into small clips, on which reciprocal attention is applied for joint motion estimation, feature alignment and feature fusion, while self attention is used for feature extraction. To enable cross-clip interactions, the video sequence is shifted for every other layer. Besides, parallel warping is used to further fuse information from neighboring frames by parallel feature warping. Experimental results on five tasks, including video super-resolution, video deblurring, video denoising, video frame interpolation and space-time video super-resolution, demonstrate that VRT outperforms the state-of-the-art methods by large margins (up to 2.16dB) on fourteen benchmark datasets. The codes are available at https://github.com/JingyunLiang/VRT.

The aim of this paper is to propose a mechanism to efficiently and explicitly model image hierarchies in the global, regional, and local range for image restoration. To achieve that, we start by analyzing two important properties of natural images including cross-scale similarity and anisotropic image features. Inspired by that, we propose the anchored stripe self-attention which achieves a good balance between the space and time complexity of self-attention and the modelling capacity beyond the regional range. Then we propose a new network architecture dubbed GRL to explicitly model image hierarchies in the Global, Regional, and Local range via anchored stripe self-attention, window self-attention, and channel attention enhanced convolution. Finally, the proposed network is applied to 7 image restoration types, covering both real and synthetic settings. The proposed method sets the new state-of-the-art for several of those. Code will be available at https://github.com/ofsoundof/GRL-Image-Restoration.git.

Abstract Self-similarity refers to the image prior widely used in image restoration algorithms that small but similar patterns tend to occur at different locations and scales. However, recent advanced deep convolutional neural network-based methods for image restoration do not take full advantage of self-similarities by relying on self-attention neural modules that only process information at the same scale. To solve this problem, we present a novel Pyramid Attention module for image restoration, which captures long-range feature correspondences from a multi-scale feature pyramid. Inspired by the fact that corruptions, such as noise or compression artifacts, drop drastically at coarser image scales, our attention module is designed to be able to borrow clean signals from their “clean” correspondences at the coarser levels. The proposed pyramid attention module is a generic building block that can be flexibly integrated into various neural architectures. Its effectiveness is validated through extensive experiments on multiple image restoration tasks: image denoising, demosaicing, compression artifact reduction, and super resolution. Without any bells and whistles, our PANet (pyramid attention module with simple network backbones) can produce state-of-the-art results with superior accuracy and visual quality. Our code is available at https://github.com/SHI-Labs/Pyramid-Attention-Networks

Video restoration aims at restoring multiple high-quality frames from multiple low-quality frames. Existing video restoration methods generally fall into two extreme cases, i.e., they either restore all frames in parallel or restore the video frame by frame in a recurrent way, which would result in different merits and drawbacks. Typically, the former has the advantage of temporal information fusion. However, it suffers from large model size and intensive memory consumption; the latter has a relatively small model size as it shares parameters across frames; however, it lacks long-range dependency modeling ability and parallelizability. In this paper, we attempt to integrate the advantages of the two cases by proposing a recurrent video restoration transformer, namely RVRT. RVRT processes local neighboring frames in parallel within a globally recurrent framework which can achieve a good trade-off between model size, effectiveness, and efficiency. Specifically, RVRT divides the video into multiple clips and uses the previously inferred clip feature to estimate the subsequent clip feature. Within each clip, different frame features are jointly updated with implicit feature aggregation. Across different clips, the guided deformable attention is designed for clip-to-clip alignment, which predicts multiple relevant locations from the whole inferred clip and aggregates their features by the attention mechanism. Extensive experiments on video super-resolution, deblurring, and denoising show that the proposed RVRT achieves state-of-the-art performance on benchmark datasets with balanced model size, testing memory and runtime.

Both Non-Local (NL) operation and sparse representation are crucial for Single Image Super-Resolution (SISR). In this paper, we investigate their combinations and propose a novel Non-Local Sparse Attention (NLSA) with dynamic sparse attention pattern. NLSA is designed to retain long-range modeling capability from NL operation while enjoying robustness and high-efficiency of sparse representation. Specifically, NLSA rectifies non-local attention with spherical locality sensitive hashing (LSH) that partitions the input space into hash buckets of related features. For every query signal, NLSA assigns a bucket to it and only computes attention within the bucket. The resulting sparse attention prevents the model from attending to locations that are noisy and less-informative, while reducing the computational cost from quadratic to asymptotic linear with respect to the spatial size. Extensive experiments validate the effectiveness and efficiency of NLSA. With a few non-local sparse attention modules, our architecture, called non-local sparse network (NLSN), reaches state-of-the-art performance for SISR quantitatively and qualitatively.

While scale-invariant modeling has substantially boosted the performance of visual recognition tasks, it remains largely under-explored in deep networks based image restoration. Naively applying those scale-invariant techniques (e.g., multi-scale testing, random-scale data augmentation) to image restoration tasks usually leads to inferior performance. In this paper, we show that properly modeling scale-invariance into neural networks can bring significant benefits to image restoration performance. Inspired from spatial-wise convolution for shift-invariance, “scale-wise convolution” is proposed to convolve across multiple scales for scale-invariance. In our scale-wise convolutional network (SCN), we first map the input image to the feature space and then build a feature pyramid representation via bi-linear down-scaling progressively. The feature pyramid is then passed to a residual network with scale-wise convolutions. The proposed scale-wise convolution learns to dynamically activate and aggregate features from different input scales in each residual building block, in order to exploit contextual information on multiple scales. In experiments, we compare the restoration accuracy and parameter efficiency among our model and many different variants of multi-scale neural networks. The proposed network with scale-wise convolution achieves superior performance in multiple image restoration tasks including image super-resolution, image denoising and image compression artifacts removal. Code and models are available at: https://github.com/ychfan/scn_sr.

Deep convolution-based single image super-resolution (SISR) networks embrace the benefits of learning from large-scale external image resources for local recovery, yet most existing works have ignored the long-range feature-wise similarities in natural images. Some recent works have successfully leveraged this intrinsic feature correlation by exploring non-local attention modules. However, none of the current deep models have studied another inherent property of images: cross-scale feature correlation. In this paper, we propose the first Cross-Scale Non-Local (CS-NL) attention module with integration into a recurrent neural network. By combining the new CS-NL prior with local and in-scale non-local priors in a powerful recurrent fusion cell, we can find more cross-scale feature correlations within a single low-resolution (LR) image. The performance of SISR is significantly improved by exhaustively integrating all possible priors. Extensive experiments demonstrate the effectiveness of the proposed CS-NL module by setting new state-of-the-arts on multiple SISR benchmarks.

Inspired by the robustness and efficiency of sparse representation in sparse coding based image restoration models, we investigate the sparsity of neurons in deep networks. Our method structurally enforces sparsity constraints upon hidden neurons. The sparsity constraints are favorable for gradient-based learning algorithms and attachable to convolution layers in various networks. Sparsity in neurons enables computation saving by only operating on non-zero components without hurting accuracy. Meanwhile, our method can magnify representation dimensionality and model capacity with negligible additional computation cost. Experiments show that sparse representation is crucial in deep neural networks for multiple image restoration tasks, including image super-resolution, image denoising, and image compression artifacts removal.

In this paper we present a new computer vision task, named video instance segmentation. The goal of this new task is simultaneous detection, segmentation and tracking of instances in videos. In words, it is the first time that the image instance segmentation problem is extended to the video domain. To facilitate research on this new task, we propose a large-scale benchmark called YouTube-VIS, which consists of 2,883 high-resolution YouTube videos, a 40-category label set and 131k high-quality instance masks. In addition, we propose a novel algorithm called MaskTrack R-CNN for this task. Our new method introduces a new tracking branch to Mask R-CNN to jointly perform the detection, segmentation and tracking tasks simultaneously. Finally, we evaluate the proposed method and several strong baselines on our new dataset. Experimental results clearly demonstrate the advantages of the proposed algorithm and reveal insight for future improvement. We believe the video instance segmentation task will motivate the community along the line of research for video understanding.

Many classic methods have shown non-local self-similarity in natural images to be an effective prior for image restoration. However, it remains unclear and challenging to make use of this intrinsic property via deep networks. In this paper, we propose a non-local recurrent network (NLRN) as the first attempt to incorporate non-local operations into a recurrent neural network (RNN) for image restoration. The main contributions of this work are: (1) Unlike existing methods that measure self-similarity in an isolated manner, the proposed non-local module can be flexibly integrated into existing deep networks for end-to-end training to capture deep feature correlation between each location and its neighborhood. (2) We fully employ the RNN structure for its parameter efficiency and allow deep feature correlation to be propagated along adjacent recurrent states. This new design boosts robustness against inaccurate correlation estimation due to severely degraded images. (3) We show that it is essential to maintain a confined neighborhood for computing deep feature correlation given degraded images. This is in contrast to existing practice that deploys the whole image. Extensive experiments on both image denoising and super-resolution tasks are conducted. Thanks to the recurrent non-local operations and correlation propagation, the proposed NLRN achieves superior results to state-of-the-art methods with many fewer parameters.

Video super-resolution (SR) aims at estimating a high-resolution (HR) video sequence from a low-resolution (LR) one. Given that deep learning has been successfully applied to the task of single image SR, which demonstrates the strong capability of neural networks for modeling spatial relation within one single image, the key challenge to conduct video SR is how to efficiently and effectively exploit the temporal dependency among consecutive LR frames other than the spatial relation. However, this remains challenging because complex motion is difficult to model and can bring detrimental effects if not handled properly. We tackle the problem of learning temporal dynamics from two aspects. First, we propose a temporal adaptive neural network that can adaptively determine the optimal scale of temporal dependency. Inspired by the Inception module in GoogLeNet [1], filters of various temporal scales are applied to the input LR sequence before their responses are adaptively aggregated, in order to fully exploit the temporal relation among consecutive LR frames. Second, we decrease the complexity of motion among neighboring frames using a spatial alignment network that can be end-to-end trained with the temporal adaptive network and has the merit of increasing the robustness to complex motion and the efficiency compared to competing image alignment methods. We provide a comprehensive evaluation of the temporal adaptation and the spatial alignment modules. We show the temporal adaptive design considerably improve SR quality over its plain counterparts, and the spatial alignment network is able to attain comparable SR performance with the sophisticated optical flow based approach, but requires much less running time. Overall our proposed model with learned temporal dynamics is shown to achieve state-of-the-art SR results in terms of not only spatial consistency but also temporal coherence on public video datasets. More information can be found in.

Video super-resolution (SR) aims to generate a high-resolution (HR) frame from multiple low-resolution (LR) frames in a local temporal window. The inter-frame temporal relation is as crucial as the intra-frame spatial relation for tackling this problem. However, how to utilize temporal information efficiently and effectively remains challenging since complex motion is difficult to model and can introduce adverse effects if not handled properly. We address this problem from two aspects. First, we propose a temporal adaptive neural network that can adaptively determine the optimal scale of temporal dependency. Filters on various temporal scales are applied to the input LR sequence before their responses are adaptively aggregated. Second, we reduce the complexity of motion between neighboring frames using a spatial alignment network which is much more robust and efficient than competing alignment methods and can be jointly trained with the temporal adaptive network in an end-to-end manner. Our proposed models with learned temporal dynamics are systematically evaluated on public video datasets and achieve state-of-the-art SR results compared with other recent video SR approaches. Both of the temporal adaptation and the spatial alignment modules are demonstrated to considerably improve SR quality over their plain counterparts.

Multi-speaker TTS trained with a general DNN has outperformed individually modelled baseline [1]. Multi-speaker DNN takes advantages of larger amount of training data from multiple speakers to find robust transformations in the hidden layers and covers more speaker variability in the output regression layer. In this paper, we propose a new approach to unsupervised speaker adaptation with multi-speaker DNN. It takes advantage of shared hidden transformation to search for the labels of unlabelled acoustic frames and the found labels are used for speaker adaption. Experimental results show that the new approach of unsupervised adaptation can achieve comparable performance with supervised adaptation both objectively and subjectively. We further extend it to cross-lingual adaptation. It can remove non-native accent and improve the naturalness while keep the same speaker's characteristics.

We have successfully proposed to use multi-speaker modelling in DNN-based TTS synthesis for improved voice quality with limited available data from a speaker. In this paper, we propose a new speaker and language factorized DNN, where speaker-specific layers are used for multi-speaker modelling, and shared layers and language-specific layers are employed for multi-language, linguistic feature transformation. Experimental results on a speech corpus of multiple speakers in both Mandarin and English show that the proposed factorized DNN can not only achieve a similar voice quality as that of a multi-speaker DNN, but also perform polyglot synthesis with a monolingual speaker's voice.

In DNN-based TTS synthesis, DNNs hidden layers can be viewed as deep transformation for linguistic features and the output layers as representation of acoustic space to regress the transformed linguistic features to acoustic parameters. The deep-layered architectures of DNN can not only represent highly-complex transformation compactly, but also take advantage of huge amount of training data. In this paper, we propose an approach to model multiple speakers TTS with a general DNN, where the same hidden layers are shared among different speakers while the output layers are composed of speaker-dependent nodes explaining the target of each speaker. The experimental results show that our approach can significantly improve the quality of synthesized speech objectively and subjectively, comparing with speech synthesized from the individual, speaker-dependent DNN-based TTS. We further transfer the hidden layers for a new speaker with limited training data and the resultant synthesized speech of the new speaker can also achieve a good quality in term of naturalness and speaker similarity.

Context-dependent deep neural network (CD-DNN) has been successfully used in large vocabulary continuous speech recognition (LVCSR). However the immense computational cost of the mini-batch based back-propagation (BP) training has become a major block to utilize massive speech data for DNN training. Previous works on BP training acceleration mainly focus on parallelization with multiple GPUs. In this paper, a novel stochastic data sweeping (SDS) framework is proposed from a different perspective to speed up DNN training with a single GPU. Part of the training data is randomly selected from the whole set and the quantity is gradually reduced at each training epoch. SDS utilizes less data in the entire process and consequently save tremendous training time. Since SDS works at data level, it is complementary to parallel training strategies and can be integrated to form a much faster training framework. Experiments showed that, combining SDS with asynchronous stochastic gradient descent (ASGD) can achieve almost 3.0 times speed-up on 2 GPUs at no loss of recognition accuracy.

Although deep neural networks (DNN) has achieved significant accuracy improvements in speech recognition, it is computationally expensive to deploy large-scale DNN in decoding due to huge number of parameters. Weights truncation and decomposition methods have been proposed to speed up decoding by exploiting the sparseness of DNN. This paper summarizes different approaches of restructuring DNN and proposes a new node pruning approach to reshape DNN for fast decoding. In this approach, hidden nodes of a fully trained DNN are pruned with certain importance function and the reshaped DNN is retuned using back-propagation. The approach requires no modification on code and can directly save computational costs during decoding. Furthermore, it is complementary to weight decomposition methods. Experiments on a switchboard task shows that, by using the proposed node-pruning approach, DNN complexity can be reduced to 37.9%. The complexity can be further reduced to 12.3% without accuracy loss when node-pruning is combined with weight decomposition.

Deep Neural Network (DNN), which can model a long-span, intricate transform compactly with a deep-layered structure, has recently been investigated for parametric TTS synthesis with a fairly large corpus (33,000 utterances) [6]. In this paper, we examine DNN TTS synthesis with a moderate size corpus of 5 hours, which is more commonly used for parametric TTS training. DNN is used to map input text features into output acoustic features (LSP, F0 and V/U). Experimental results show that DNN can outperform the conventional HMM, which is trained in ML first and then refined by MGE. Both objective and subjective measures indicate that DNN can synthesize speech better than HMM-based baseline. The improvement is mainly on the prosody, i.e., the RMSE of natural and generated F0 trajectories by DNN is improved by 2 Hz. This benefit is likely from the key characteristics of DNN, which can exploit feature correlations, e.g., between F0 and spectrum, without using a more restricted, e.g. diagonal Gaussian probability family. Our experimental results also show: the layer-wise BP pre-training can drive weights to a better starting point than random initialization and result in a more effective DNN; state boundary info is important for training DNN to yield better synthesized speech; and a hyperbolic tangent activation function in DNN hidden layers yields faster convergence than a sigmoidal one.

Due to great success of deep learning in speech recognition, there has been interest of applying deep learning to speaker verification. Previous investigations usually focus on using deep neural network as new classifiers or to extract speaker dependent features. They are either not compatible with existing speaker verification approaches, or not able to achieve significant performance gain in large scale tasks. Also, all the previous approaches have not addressed the issue of how to make use of extra unsupervised data. This paper proposes a novel feature engineering approach within the deep learning framework for speaker verification. Hidden layer output of deep neural network or deep belief network trained on large amount of speech recognition data are extracted as deep features. These features are then used in a Tandem fashion or concatenated with the original acoustic features for GMM-UBM speaker verification. The proposed approach can make use of large amount of existing speech recognition data without speaker labels and is easy to be combined with other mature classification approaches. Experiments on the core condition of NIST 2006 SRE showed that, in a text independent task, the proposed approach can achieve 12.8% relative EER improvement compared to the standard GMM-UBM systems. In addition, text-dependent speaker verification experiments were also performed and yielded similar significant gain.

Feed-forward, Deep neural networks (DNN)-based text-tospeech (TTS) systems have been recently shown to outperform decision-tree clustered context-dependent HMM TTS systems [1, 4]. However, the long time span contextual effect in a speech utterance is still not easy to accommodate, due to the intrinsic, feed-forward nature in DNN-based modeling. Also, to synthesize a smooth speech trajectory, the dynamic features are commonly used to constrain speech parameter trajectory generation in HMM-based TTS [2]. In this paper, Recurrent Neural Networks (RNNs) with Bidirectional Long Short Term Memory (BLSTM) cells are adopted to capture the correlation or co-occurrence information between any two instants in a speech utterance for parametric TTS synthesis. Experimental results show that a hybrid system of DNN and BLSTM-RNN, i.e., lower hidden layers with a feed-forward structure which is cascaded with upper hidden layers with a bidirectional RNN structure of LSTM, can outperform either the conventional, decision tree-based HMM, or a DNN TTS system, both objectively and subjectively. The speech trajectory generated by the BLSTM-RNN TTS is fairly smooth and no dynamic constraints are needed.

Neural network (NN) based voice conversion, which employs a nonlinear function to map the features from a source to a target speaker, has been shown to outperform GMM-based voice conversion approach [4-7]. However, there are still limitations to be overcome in NN-based voice conversion, e.g. NN is trained on a Frame Error (FE) minimization criterion and the corresponding weights are adjusted to minimize the error squares over the whole source-target, stereo training data set. In this paper, we use the idea of sentence optimization based, minimum generation error (MGE) training in HMM-based TTS synthesis, and modify the FE minimization to Sequence Error (SE) minimization in NN training for voice conversion. The conversion error over a training sentence from a source speaker to a target speaker is minimized via a gradient descent-based, back propagation (BP) procedure. Experimental results show that the speech converted by the NN, which is first trained with frame error minimization and then refined with sequence error minimization, sounds subjectively better than the converted speech by NN trained with frame error minimization only. Scores on both naturalness and similarity to the target speaker are improved. Index Terms: voice conversion, neural network, pre-training, sequence error minimization

Haptic voice recognition (HVR) is a multi-modal text entry method for smart mobile devices. It employs haptic events generated by speakers during speaking to achieve better efficiency and robustness for automatic speech recognition. This paper describes the detailed design of the 2012 SJTU submission for the HVR Grand Challenge. During the design, a new perplexity metric using conditional entropy is proposed to evaluate the potential search space reduction of a haptic event without speech input. A number of new haptic events are evaluated both theoretically and experimentally in detail. The final submission system uses the haptic event of initial letter plus final letter and reduces word error rate by 76% compared to the baseline initial letter event.