Due to the recent advances of natural language processing, several works have applied the pre-trained masked language model (MLM) of BERT to the post-correction of speech recognition. However, existing pre-trained models only consider the semantic correction while the phonetic features of words is neglected. The semantic-only post-correction will consequently decrease the performance since homophonic errors are fairly common in Chinese ASR. In this paper, we proposed a novel approach to collectively exploit the contextualized representation and the phonetic information between the error and its replacing candidates to alleviate the error rate of Chinese ASR. Our experiment results on real world speech recognition datasets showed that our proposed method has evidently lower CER than the baseline model, which utilized a pre-trained BERT MLM as the corrector.
With increasing popularity of social media platforms hate speech is emerging as a major concern, where it expresses abusive speech that targets specific group characteristics, such as gender, religion or ethnicity to spread violence. Earlier people use to verbally deliver hate speeches but now with the expansion of technology, some people are deliberately using social media platforms to spread hate by posting, sharing, commenting, etc. Whether it is Christchurch mosque shootings or hate crimes against Asians in west, it has been observed that the convicts are very much influenced from hate text present online. Even though AI systems are in place to flag such text but one of the key challenges is to reduce the false positive rate (marking non hate as hate), so that these systems can detect hate speech without undermining the freedom of expression. In this paper, we use ETHOS hate speech detection dataset and analyze the performance of hate speech detection classifier by replacing or integrating the word embeddings (fastText (FT), GloVe (GV) or FT + GV) with static BERT embeddings (BE). With the extensive experimental trails it is observed that the neural network performed better with static BE compared to using FT, GV or FT + GV as word embeddings. In comparison to fine-tuned BERT, one metric that significantly improved is specificity.
Recent advances in the design of neural network architectures, in particular those specialized in modeling sequences, have provided significant improvements in speech separation performance. In this work, we propose to use a bio-inspired architecture called Fully Recurrent Convolutional Neural Network (FRCNN) to solve the separation task. This model contains bottom-up, top-down and lateral connections to fuse information processed at various time-scales represented by \textit{stages}. In contrast to the traditional approach updating stages in parallel, we propose to first update the stages one by one in the bottom-up direction, then fuse information from adjacent stages simultaneously and finally fuse information from all stages to the bottom stage together. Experiments showed that this asynchronous updating scheme achieved significantly better results with much fewer parameters than the traditional synchronous updating scheme. In addition, the proposed model achieved good balance between speech separation accuracy and computational efficiency as compared to other state-of-the-art models on three benchmark datasets.
This paper describes the Microsoft end-to-end neural text to speech (TTS) system: DelightfulTTS for Blizzard Challenge 2021. The goal of this challenge is to synthesize natural and high-quality speech from text, and we approach this goal in two perspectives: The first is to directly model and generate waveform in 48 kHz sampling rate, which brings higher perception quality than previous systems with 16 kHz or 24 kHz sampling rate; The second is to model the variation information in speech through a systematic design, which improves the prosody and naturalness. Specifically, for 48 kHz modeling, we predict 16 kHz mel-spectrogram in acoustic model, and propose a vocoder called HiFiNet to directly generate 48 kHz waveform from predicted 16 kHz mel-spectrogram, which can better trade off training efficiency, modelling stability and voice quality. We model variation information systematically from both explicit (speaker ID, language ID, pitch and duration) and implicit (utterance-level and phoneme-level prosody) perspectives: 1) For speaker and language ID, we use lookup embedding in training and inference; 2) For pitch and duration, we extract the values from paired text-speech data in training and use two predictors to predict the values in inference; 3) For utterance-level and phoneme-level prosody, we use two reference encoders to extract the values in training, and use two separate predictors to predict the values in inference. Additionally, we introduce an improved Conformer block to better model the local and global dependency in acoustic model. For task SH1, DelightfulTTS achieves 4.17 mean score in MOS test and 4.35 in SMOS test, which indicates the effectiveness of our proposed system
Advances in deep learning have introduced a new wave of voice synthesis tools, capable of producing audio that sounds as if spoken by a target speaker. If successful, such tools in the wrong hands will enable a range of powerful attacks against both humans and software systems (aka machines). This paper documents efforts and findings from a comprehensive experimental study on the impact of deep-learning based speech synthesis attacks on both human listeners and machines such as speaker recognition and voice-signin systems. We find that both humans and machines can be reliably fooled by synthetic speech and that existing defenses against synthesized speech fall short. These findings highlight the need to raise awareness and develop new protections against synthetic speech for both humans and machines.
The widespread adoption of speech-based online services raises security and privacy concerns regarding the data that they use and share. If the data were compromised, attackers could exploit user speech to bypass speaker verification systems or even impersonate users. To mitigate this, we propose DeID-VC, a speaker de-identification system that converts a real speaker to pseudo speakers, thus removing or obfuscating the speaker-dependent attributes from a spoken voice. The key components of DeID-VC include a Variational Autoencoder (VAE) based Pseudo Speaker Generator (PSG) and a voice conversion Autoencoder (AE) under zero-shot settings. With the help of PSG, DeID-VC can assign unique pseudo speakers at speaker level or even at utterance level. Also, two novel learning objectives are added to bridge the gap between training and inference of zero-shot voice conversion. We present our experimental results with word error rate (WER) and equal error rate (EER), along with three subjective metrics to evaluate the generated output of DeID-VC. The result shows that our method substantially improved intelligibility (WER 10% lower) and de-identification effectiveness (EER 5% higher) compared to our baseline. Code and listening demo: https://github.com/a43992899/DeID-VC
Real-Time Magnetic resonance imaging (rtMRI) of the midsagittal plane of the mouth is of interest for speech production research. In this work, we focus on estimating utterance level rtMRI video from the spoken phoneme sequence. We obtain time-aligned phonemes from forced alignment, to obtain frame-level phoneme sequences which are aligned with rtMRI frames. We propose a sequence-to-sequence learning model with a transformer phoneme encoder and convolutional frame decoder. We then modify the learning by using intermediary features obtained from sampling from a pretrained phoneme-conditioned variational autoencoder (CVAE). We train on 8 subjects in a subject-specific manner and demonstrate the performance with a subjective test. We also use an auxiliary task of air tissue boundary (ATB) segmentation to obtain the objective scores on the proposed models. We show that the proposed method is able to generate realistic rtMRI video for unseen utterances, and adding CVAE is beneficial for learning the sequence-to-sequence mapping for subjects where the mapping is hard to learn.
We present a method for cross-lingual training an ASR system using absolutely no transcribed training data from the target language, and with no phonetic knowledge of the language in question. Our approach uses a novel application of a decipherment algorithm, which operates given only unpaired speech and text data from the target language. We apply this decipherment to phone sequences generated by a universal phone recogniser trained on out-of-language speech corpora, which we follow with flat-start semi-supervised training to obtain an acoustic model for the new language. To the best of our knowledge, this is the first practical approach to zero-resource cross-lingual ASR which does not rely on any hand-crafted phonetic information. We carry out experiments on read speech from the GlobalPhone corpus, and show that it is possible to learn a decipherment model on just 20 minutes of data from the target language. When used to generate pseudo-labels for semi-supervised training, we obtain WERs that range from 25% to just 5% absolute worse than the equivalent fully supervised models trained on the same data.
The perceptual task of speech quality assessment (SQA) is a challenging task for machines to do. Objective SQA methods that rely on the availability of the corresponding clean reference have been the primary go-to approaches for SQA. Clearly, these methods fail in real-world scenarios where the ground truth clean references are not available. In recent years, non-intrusive methods that train neural networks to predict ratings or scores have attracted much attention, but they suffer from several shortcomings such as lack of robustness, reliance on labeled data for training and so on. In this work, we propose a new direction for speech quality assessment. Inspired by human's innate ability to compare and assess the quality of speech signals even when they have non-matching contents, we propose a novel framework that predicts a subjective relative quality score for the given speech signal with respect to any provided reference without using any subjective data. We show that neural networks trained using our framework produce scores that correlate well with subjective mean opinion scores (MOS) and are also competitive to methods such as DNSMOS, which explicitly relies on MOS from humans for training networks. Moreover, our method also provides a natural way to embed quality-related information in neural networks, which we show is helpful for downstream tasks such as speech enhancement.
Deep neural networks, such as the Deep-FSMN, have been widely studied for keyword spotting (KWS) applications while suffering expensive computation and storage. Therefore, network compression technologies like binarization are studied to deploy KWS models on edge. In this paper, we present a strong yet efficient binary neural network for KWS, namely BiFSMNv2, pushing it to the real-network accuracy performance. First, we present a Dual-scale Thinnable 1-bit-Architecture to recover the representation capability of the binarized computation units by dual-scale activation binarization and liberate the speedup potential from an overall architecture perspective. Second, we also construct a Frequency Independent Distillation scheme for KWS binarization-aware training, which distills the high and low-frequency components independently to mitigate the information mismatch between full-precision and binarized representations. Moreover, we implement BiFSMNv2 on ARMv8 real-world hardware with a novel Fast Bitwise Computation Kernel, which is proposed to fully utilize registers and increase instruction throughput. Comprehensive experiments show our BiFSMNv2 outperforms existing binary networks for KWS by convincing margins across different datasets and even achieves comparable accuracy with the full-precision networks (e.g., only 1.59% drop on Speech Commands V1-12). We highlight that benefiting from the compact architecture and optimized hardware kernel, BiFSMNv2 can achieve an impressive 25.1x speedup and 20.2x storage-saving on edge hardware.