Autoregressive neural vocoders have achieved outstanding performance and are widely used in speech synthesis tasks such as text-to-speech and voice conversion. An autoregressive vocoder predicts a sample at some time step conditioned on those at previous time steps. Though it can generate highly natural human speech, the iterative generation inevitably makes the synthesis time proportional to the utterance length, leading to low efficiency. Many works were dedicated to generating the whole speech time sequence in parallel and then proposed GAN-based, flow-based, and score-based vocoders. This paper proposed a new thought for the autoregressive generation. Instead of iteratively predicting samples in a time sequence, the proposed model performs frequency-wise autoregressive generation (FAR) and bit-wise autoregressive generation (BAR) to synthesize speech. In FAR, a speech utterance is first split into different frequency subbands. The proposed model generates a subband conditioned on the previously generated one. A full-band speech can then be reconstructed from these generated subbands. Similarly, in BAR, an 8-bit quantized signal is generated iteratively from the first bit. By redesigning the autoregressive method to compute in domains other than the time domain, the number of iterations in the proposed model is no longer proportional to the utterance's length but to the number of subbands/bits. The inference efficiency is hence significantly increased. Besides, a post-filter is employed to sample audio signals from output posteriors, and its training objective is designed based on the characteristics of the proposed autoregressive methods. The experimental results show that the proposed model can synthesize speech faster than real-time without GPU acceleration. Compared with the baseline autoregressive and non-autoregressive vocoders, the proposed model achieves better MUSHRA results and shows good generalization ability while synthesizing 44 kHz speech or utterances from unseen speakers.

Automatic speaker verification (ASV), one of the most important technology for biometric identification, has been widely adopted in security-critical applications. However, ASV is seriously vulnerable to recently emerged adversarial attacks, yet effective counter-measures against them are limited. In this paper, we adopt neural vocoders to spot adversarial samples for ASV. We use the neural vocoder to re-synthesize audio and find that the difference between the ASV scores for the original and re-synthesized audio is a good indicator for discrimination between genuine and adversarial samples. This effort is, to the best of our knowledge, among the first to pursue such a technical direction for detecting time-domain adversarial samples for ASV, and hence there is a lack of established baselines for comparison. Consequently, we implement the Griffin-Lim algorithm as the detection baseline. The proposed approach achieves effective detection performance that outperforms the baselines in all the settings. We also show that the neural vocoder adopted in the detection framework is dataset-independent. Our codes will be made open-source for future works to do fair comparison1.

End-to-end spoken language understanding (SLU) predicts intent directly from audio using a single model. It promises to improve the performance of assistant systems by leveraging acoustic information lost in the intermediate textual representation and preventing cascading errors from Automatic Speech Recognition (ASR). Further, having one unified model has efficiency advantages when deploying assistant systems on-device. However, the limited number of public audio datasets with semantic parse labels hinders the research progress in this area. In this paper, we release the Spoken Task-Oriented semantic Parsing (STOP) dataset1, the largest and most complex SLU dataset publicly available. Additionally, we define low-resource splits to establish a benchmark for improving SLU when limited labeled data is available. Furthermore, in addition to the human-recorded audio, we are releasing a TTS-generated versions to benchmark the performance for low-resource and domain adaptation of end-to-end SLU systems.

Terahertz (THz) communication with ultra-wide available spectrum is a promising technique that can achieve the stringent requirement of high data rate in the next-generation wireless networks, yet its severe propagation attenuation significantly hinders its implementation in practice. Finding beam directions for a large-scale antenna array to effectively overcome severe propagation attenuation of THz signals is a pressing need. This paper proposes a novel approach of federated deep reinforcement learning (FDRL) to swiftly perform THz-beam search for multiple base stations (BSs) coordinated by an edge server in a cellular network. All the BSs conduct deep deterministic policy gradient (DDPG)-based DRL to obtain THz beamforming policy with limited channel state information (CSI). They update their DDPG models with hidden information in order to mitigate inter-cell interference. We demonstrate that the cell network can achieve higher throughput as more THz CSI and hidden neurons of DDPG are adopted. We also show that FDRL with partial model update is able to nearly achieve the same performance of FDRL with full model update, which indicates an effective means to reduce communication load between the edge server and the BSs by partial model uploading. Moreover, the proposed FDRL outperforms conventional non-learning-based and existing non-FDRL benchmark optimization methods.

ASRhas been shown to achieve great performance recently. However, most of them rely on massive paired data, which is not feasible for low-resource languages worldwide. This paper investigates how to learn directly from unpaired phone sequences and speech utterances. We design a two-stage iterative framework. GAN training is adopted in the first stage to find the mapping relationship between unpaired speech and phone sequence. In the second stage, another HMM model is introduced to train from the generator’s output, which boosts the performance and provides a better segmentation for the next iteration. In the experiment, we first investigate different choices of model designs. Thenwe compare the framework to different types of baselines: (i) supervised methods (ii) acoustic unit discovery based methods (iii) methods learning from unpaired data. Our framework performs consistently better than all acoustic unit discovery methods and previous methods learning from unpaired data based on the TIMIT dataset.

The few-shot multi-speaker multi-style voice cloning task is to synthesize utterances with voice and speaking style similar to a reference speaker given only a few reference samples. In this work, we investigate different speaker representations and proposed to integrate pretrained and learnable speaker representations. Among different types of embeddings, the embedding pretrained by voice conversion achieves the best performance. The FastSpeech 2 model combined with both pretrained and learnable speaker representations shows great generalization ability on few-shot speakers and achieved 2nd place in the one-shot track of the ICASSP 2021 M2VoC challenge.

We present Mockingjay as a new speech representation learning approach, where bidirectional Transformer encoders are pre-trained on a large amount of unlabeled speech. Previous speech representation methods learn through conditioning on past frames and predicting information about future frames. Whereas Mockingjay is designed to predict the current frame through jointly conditioning on both past and future contexts. The Mockingjay representation improves performance for a wide range of downstream tasks, including phoneme classification, speaker recognition, and sentiment classification on spoken content, while outperforming other approaches. Mockingjay is empirically powerful and can be fine-tuned with downstream models, with only 2 epochs we further improve performance dramatically. In a low resource setting with only 0.1% of labeled data, we outperform the result of Mel-features that uses all 100% labeled data.

In this paper, we propose WG-WaveNet, a fast, lightweight, and high-quality waveform generation model. WG-WaveNet is composed of a compact flow-based model and a post-filter. The two components are jointly trained by maximizing the likelihood of the training data and optimizing loss functions on the frequency domains. As we design a flow-based model that is heavily compressed, the proposed model requires much less computational resources compared to other waveform generation models during both training and inference time; even though the model is highly compressed, the post-filter maintains the quality of generated waveform. Our PyTorch implementation can be trained using less than 8 GB GPU memory and generates audio samples at a rate of more than 960 kHz on an NVIDIA 1080Ti GPU. Furthermore, even if synthesizing on a CPU, we show that the proposed method is capable of generating 44.1 kHz speech waveform 1.2 times faster than real-time. Experiments also show that the quality of generated audio is comparable to those of other methods. Audio samples are publicly available online.

We present an unsupervised end-to-end training scheme where we discover discrete subword units from speech without using any labels. The discrete subword units are learned under an ASR-TTS autoencoder reconstruction setting, where an ASR-Encoder is trained to discover a set of common linguistic units given a variety of speakers, and a TTS-Decoder trained to project the discovered units back to the designated speech. We propose a discrete encoding method, Multilabel-Binary Vectors (MBV), to make the ASR-TTS autoencoder differentiable. We found that the proposed encoding method offers automatic extraction of speech content from speaker style, and is sufficient to cover full linguistic content in a given language. Therefore, the TTS-Decoder can synthesize speech with the same content as the input of ASR-Encoder but with different speaker characteristics, which achieves voice conversion (VC). We further improve the quality of VC using adversarial training, where we train a TTS-Patcher that augments the output of TTS-Decoder. Objective and subjective evaluations show that the proposed approach offers strong VC results as it eliminates speaker identity while preserving content within speech. In the ZeroSpeech 2019 Challenge, we achieved outstanding performance in terms of low bitrate.

Speaking rate refers to the average number of phonemes within some unit time, while the rhythmic patterns refer to duration distributions for realizations of different phonemes within different phonetic structures. Both are key components of prosody in speech, which is different for different speakers. Models like cycle-consistent adversarial network (Cycle-GAN) and variational auto-encoder (VAE) have been successfully applied to voice conversion tasks without parallel data. However, due to the neural network architectures and feature vectors chosen for these approaches, the length of the predicted utterance has to be fixed to that of the input utterance, which limits the flexibility in mimicking the speaking rates and rhythmic patterns for the target speaker. On the other hand, sequence-to-sequence learning model was used to remove the above length constraint, but parallel training data are needed. In this paper, we propose an approach utilizing sequence-to-sequence model trained with unsupervised Cycle-GAN to perform the transformation between the phoneme posteriorgram sequences for different speakers. In this way, the length constraint mentioned above is removed to offer rhythm-flexible voice conversion without requiring parallel data. Preliminary evaluation on two datasets showed very encouraging results.

This study examines the effects of heavy rainfall on the Kuroshio surface salinity east of Taiwan. The main focus of this study is on the abnormal hydrologic characteristics of the Kuroshio after heavy rainfall. Data of sea surface salinity (SSS) are derived from R/V Ocean Researcher I (R/V OR1) 1012 cruise and 8 glider cruises; data of rain rate from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) which helps to figure out the change of SSS of the Kuroshio from field surveys and glider datasets. The results show a good correlation between the rain rate and minimum SSS with a coefficient of determination of 0.74. The rainfall causes the SSS of the Kuroshio surface water drops 0.176 psu per 1 mm/hr rain rate. This underscore the importance of hydrologic characteristic fluctuations must be taken into account for seasonal or long-term variability study.

This paper proposes a niching scheme, the dependency structure matrix restricted tournament replacement (DSMRTR). The restricted tournament replacement (RTR) is a well-known niching scheme in the field of estimation of distribution algorithms (EDAs). However, RTR induces spurious dependencies among variables, which impair the performance of EDAs. This paper utilizes building-block-wise distances to define a new distance metric, the one-niche distance. For those EDAs which provide explicit linkage information, the one-niche distances can be directly incorporated into RTR. For EDAs without such information, DSMRTR constructs a dependency structure matrix via the differential mutual complement to estimate the one-niche distances. Empirical results show that DSMRTR induces fewer spurious dependencies than RTR does while maintaining enough diversity for EDAs.

Introduction to Computer Science (ICS) is a compulsory course for freshmen of the electrical engineering (EE) department at our institute, which lays the foundation for more advanced courses. To arouse the interest and learning initiative of students, an optional “Special Practical Experiment in Computer Science (SPECS)” practice was introduced in 2010, emphasizing the integration of concepts with hands-on practice. Questionnaires on teacher evaluation were given to students nearing the end of the ICS course to evaluate their satisfaction with the teaching and their perception of the benefits derived from ICS. Participants and non-participants of SPECS were compared according to their post-SPECS class attendance and performance in ICS. The academic scores of SPECS participants were significantly higher than non-participants (80.16±5.33 vs. 72.79±10.39 in 2010, p<;0.05; 81.35±4.30 vs. 75.52±6.35 in 2011, p<;0.001). The SPECS not only improved students' satisfaction with teaching, but also their enthusiasm and performance in the ICS course.

In the field of genetic algorithms (GAs), some researches on overlapping building blocks (BBs) have been proposed. To further study on overlapping BBs, we need to measure the performance of an algorithm to solve problems with over-lap among subproblems. Several test problems have been proposed, but the controllability over the degree of overlapping is not yet fully satisfactory. Our new test problem is designed with full controllability of overlapping as well as conflict, a specific type of overlap, among BBs. We present a framework for building the structure of this problem in this paper. Some model-building GAs are tested by the proposed problem. This test problem can be applied to further researches on overlapping and conflicting BBs.

As the web becomes the platform of choice for execution of more complex applications, a growing portion of computation is handed off by developers to the client side to reduce network traffic and improve application responsiveness. Therefore, the client-side component, often written in JavaScript, is becoming larger and more compute-intensive, increasing the demand for high performance JavaScript execution. This has led to many recent efforts to improve the performance of JavaScript engines in the web browsers. Furthermore, considering the wide-spread deployment of multi-cores in today's computing systems, exploiting parallelism in these applications is a promising approach to meet their performance requirement. However, JavaScript has traditionally been treated as a sequential language with no support for multithreading, limiting its potential to make use of the extra computing power in multicore systems. In this work, to exploit hardware concurrency while retaining traditional sequential programming model, we develop ParaScript, an automatic runtime parallelization system for JavaScript applications on the client's browser. First, we propose an optimistic runtime scheme for identifying parallelizable regions, generating the parallel code on-the-fly, and speculatively executing it. Second, we introduce an ultra-lightweight software speculation mechanism to manage parallel execution. This speculation engine consists of a selective checkpointing scheme and a novel runtime dependence detection mechanism based on reference counting and range-based array conflict detection. Our system is able to achieve an average of 2.18× speedup over the Firefox browser using 8 threads on commodity multi-core systems, while performing all required analyses and conflict detection dynamically at runtime.

Multicore designs have emerged as the mainstream design paradigm for the microprocessor industry. Unfortunately, providing multiple cores does not directly translate into performance for most applications. The industry has already fallen short of the decades-old performance trend of doubling performance every 18 months. An attractive approach for exploiting multiple cores is to rely on tools, both compilers and runtime optimizers, to automatically extract threads from sequential applications. However, despite decades of research on automatic parallelization, most techniques are only effective in the scientific and data parallel domains where array dominated codes can be precisely analyzed by the compiler. Thread-level speculation offers the opportunity to expand parallelization to general-purpose programs, but at the cost of expensive hardware support. In this paper, we focus on providing low-overhead software support for exploiting speculative parallelism. We propose STMlite, a light-weight software transactional memory model that is customized to facilitate profile-guided automatic loop parallelization. STMlite eliminates a considerable amount of checking and locking overhead in conventional software transactional memory models by decoupling the commit phase from main transaction execution. Further, strong atomicity requirements for generic transactional memories are unnecessary within a stylized automatic parallelization framework. STMlite enables sequential applications to extract meaningful performance gains on commodity multicore hardware.