Permutation Invariant Training Research Articles

Multi-target direction-of-arrival estimation of deep models with frame-level permutation invariant training in marine acoustic environment.

Direction-of-arrival (DoA) estimation is an important part in sonar signal processing, providing a reliable foundation for tasks, such as underwater object detection and tracking. Although the deep learning model has powerful data fitting capabilities, accurately estimating the orientation of multiple targets with a single model remains a challenging task. To address this challenge, we enhance the permutation invariant training (PIT) technique and propose two different types of methods: multi-group classification with PIT (MC-PIT) and multi-group regression with PIT (MR-PIT). These two frame-level PIT schemes utilize a single model for both training and testing in multi-target scenarios. Furthermore, we evaluate the performance of MR-PIT and MC-PIT with different network backbones and demonstrate that the frame-level PIT has excellent portability. Compared with the model trained with the general multi-label strategy, simulation experiments show that our proposed methods have better multi-target DoA estimation performance. Finally, when the array configuration of simulated and recorded data are consistent, the model with frame-level PIT can achieve good performance on recorded data even only trained on simulation data.

ONLINE BINAURAL SPEECH SEPARATION OF MOVING SPEAKERS WITH A WAVESPLIT NETWORK.

Binaural speech separation in real-world scenarios often involves moving speakers. Most current speech separation methods use utterance-level permutation invariant training (u-PIT) for training. In inference time, however, the order of outputs can be inconsistent over time particularly in long-form speech separation. This situation which is referred to as the speaker swap problem is even more problematic when speakers constantly move in space and therefore poses a challenge for consistent placement of speakers in output channels. Here, we describe a real-time binaural speech separation model based on a Wavesplit network to mitigate the speaker swap problem for moving speaker separation. Our model computes a speaker embedding for each speaker at each time frame from the mixed audio, aggregates embeddings using online clustering, and uses cluster centroids as speaker profiles to track each speaker throughout the long duration. Experimental results on reverberant, long-form moving multitalker speech separation show that the proposed method is less prone to speaker swap and achieves comparable performance with u-PIT based models with ground truth tracking in both separation accuracy and preserving the interaural cues.

Single-channel speech separation using soft-minimum permutation invariant training

The goal of speech separation is to extract multiple speech sources from a single microphone recording. Recently, with the advancement of deep learning and availability of large datasets, speech separation has been formulated as a supervised learning problem. These approaches aim to learn discriminative patterns of speech, speakers, and background noise using a supervised learning algorithm, typically a deep neural network. A long-lasting problem in supervised speech separation is finding the correct label for each separated speech signal, referred to as label permutation ambiguity. Permutation ambiguity refers to the problem of determining the output-label assignment between the separated sources and the available single-speaker speech labels. Finding the best output-label assignment is required for calculation of separation error, which is later used for updating parameters of the model. Recently, Permutation Invariant Training (PIT) has been shown to be a promising solution in handling the label ambiguity problem. However, the overconfident choice of the output-label assignment by PIT results in a sub-optimal trained model. In this work, we propose a probabilistic optimization framework to address the inefficiency of PIT in finding the best output-label assignment. Our proposed method entitled trainable Soft-minimum PIT is then employed on the same Long-Short Term Memory (LSTM) architecture used in Permutation Invariant Training (PIT) speech separation method. The results of our experiments show that the proposed method outperforms conventional PIT speech separation significantly (p-value <0.01) by +1dB in Signal to Distortion Ratio (SDR) and +1.5dB in Signal to Interference Ratio (SIR).

Segment-Less Continuous Speech Separation of Meetings: Training and Evaluation Criteria

Continuous Speech Separation (CSS) has been proposed to address speech overlaps during the analysis of realistic meeting-like conversations by eliminating any overlaps before further processing. CSS separates a recording of arbitrarily many speakers into a small number of overlap-free output channels, where each output channel may contain speech of multiple speakers. Often, a separation model is trained with Utterance-level Permutation Invariant Training (uPIT), which exclusively maps a speaker to an output channel, and applied in a sliding window approach called stitching. Recently, we introduced an alternative training scheme called Graph-PIT that teaches the separator to produce a speaker-shared output channel format without stitching. It can handle an arbitrary number of speakers as long as the number of overlapping speakers is never larger than the number of output channels. Models trained in this way are able to perform segment-less CSS, i.e., without stitching, and achieve comparable and often better separation quality than the conventional CSS with uPIT and stitching. In this contribution, we further investigate the Graph-PIT training scheme. We show in extended experiments that Graph-PIT also works in challenging reverberant conditions. We simplify the training schedule for Graph-PIT with the recently proposed Source Aggregated Signal-to-Distortion Ratio (SA-SDR) loss, which eliminates unfavorable properties of the previously used A-SDR loss to enable training with Graph-PIT from scratch. Furthermore, we introduce novel signal-level evaluation metrics for meeting scenarios, namely the source-aggregated scale- and convolution-invariant Signal-to-Distortion Ratio (SA-SI-SDR and SA-CI-SDR), which are generalizations of the commonly used SDR-based metrics for the CSS case.

Single-Channel Speech Separation Using Soft-Minimum Permutation Invariant Training

The goal of speech separation is to extract multiple speech sources from a single microphone recording. Recently, with the advancement of deep learning and availability of large datasets, speech separation has been formulated as a supervised learning problem. These approaches aim to learn discriminative patterns of speech, speakers, and background noise using a supervised learning algorithm, typically a deep neural network. A long-lasting problem in supervised speech separation is finding the correct label for each separated speech signal, referred to as label permutation ambiguity. Permutation ambiguity refers to the problem of determining the output-label assignment between the separated sources and the available single-speaker speech labels. Finding the best output-label assignment is required for calculation of separation error, which is later used for updating parameters of the model. Recently, Permutation Invariant Training (PIT) has been shown to be a promising solution in handling the label ambiguity problem. However, the overconfident choice of the output-label assignment by PIT results in a sub-optimal trained model. In this work, we propose a probabilistic optimization framework to address the inefficiency of PIT in finding the best output-label assignment. Our proposed method entitled trainable Soft-minimum PIT is then employed on the same Long-Short Term Memory (LSTM) architecture used in Permutation Invariant Training (PIT) speech separation method. The results of our experiments show that the proposed method outperforms conventional PIT speech separation significantly (p-value $ < 0.01$) by +1dB in Signal to Distortion Ratio (SDR) and +1.5dB in Signal to Interference Ratio (SIR).

Multi-Channel Talker-Independent Speaker Separation Through Location-Based Training

Permutation ambiguity is a crucial issue for deep learning based talker-independent speaker separation. Deep clustering and permutation invariant training (PIT) have been widely used to address the permutation ambiguity problem in monaural scenarios. Although both approaches have been extended to multi-microphone scenarios, we believe that the permutation ambiguity problem can be naturally avoided by leveraging the spatial relations of multiple speakers. In this study, we present location-based training (LBT), a new approach to achieve talker independency in multi-channel speaker separation. Unlike PIT that examines all possible permutations, LBT assigns speakers according to their positions in physical space. With a linear training complexity to the number of concurrent speakers, LBT is computationally much more efficient than PIT with a factorial complexity, particularly when a large number of overlapping speakers needs to be separated. Specifically, we propose two training criteria: azimuth-based and distance-based training, using speaker azimuths and distances relative to a microphone array, respectively. Evaluation results show that LBT significantly outperforms PIT on two-speaker and three-speaker mixtures with different array geometries and in various acoustic conditions. In addition, we propose a joint training strategy to integrate azimuth-based and distance-based training, which further improves separation performance.

MixCycle: Unsupervised Speech Separation via Cyclic Mixture Permutation Invariant Training

We introduce two unsupervised source separation methods, which involve self-supervised training from single-channel two-source speech mixtures. Our first method, mixture permutation invariant training (MixPIT), enables learning a neural network model which separates the underlying sources via a challenging proxy task without supervision from the reference sources. Our second method, cyclic mixture permutation invariant training (MixCycle), uses MixPIT as a building block in a cyclic fashion for continuous learning. MixCycle gradually converts the problem from separating mixtures of mixtures into separating single mixtures. We compare our methods to common supervised and unsupervised baselines: permutation invariant training with dynamic mixing (PIT-DM) and mixture invariant training (MixIT). We show that MixCycle outperforms MixIT and reaches a performance level very close to the supervised baseline (PIT-DM) while circumventing the over-separation issue of MixIT. Also, we propose a self-evaluation technique inspired by MixCycle that estimates model performance without utilizing any reference sources. We show that it yields results consistent with an evaluation on reference sources (LibriMix) and also with an informal listening test conducted on a real-life mixtures dataset (REAL-M).

A dual-stream deep attractor network with multi-domain learning for speech dereverberation and separation

Deep attractor networks (DANs) perform speech separation with discriminative embeddings and speaker attractors. Compared with methods based on the permutation invariant training (PIT), DANs define a deep embedding space and deliver a more elaborate representation on each time–frequency (T–F) bin. However, it has been observed that the DANs achieve limited improvement on the signal quality if directly deployed in a reverberant environment. Following the success of time-domain separation networks on the clean mixture speech, we propose a dual-stream DAN with multi-domain learning to efficiently perform both dereverberation and separation tasks under the condition of variable numbers of speakers. The speaker encoding stream (SES) of the dual-stream DAN is trained to model the speaker information in the embedding space defined with the Fourier transform kernels. The speech decoding stream (SDS) accepts speaker attractors from the SES and learns to estimate the early component of the sound in the time domain. Meanwhile, additional clustering losses are used to bridge the gap between the oracle and the estimated attractors. Experiments were conducted on the Spatialized Multi-Speaker Wall Street Journal (SMS-WSJ) dataset. After comparing with the anechoic and reverberant signals, the early component was chosen as the learning targets. The experimental results demonstrated that the dual-stream DAN achieved scale-invariant source-to-distortion ratio (SI-SDR) improvement of 9.8∕7.5 dB on the reverberant 2-/3-speaker evaluation set, exceeding the baseline DAN and convolutional time-domain audio separation network (Conv-TasNet) by 2.0∕0.7 dB and 1.0∕0.5 dB, respectively.

Multichannel environmental sound segmentation

This paper proposes a multichannel environmental sound segmentation method. Environmental sound segmentation is an integrated method to achieve sound source localization, sound source separation and classification, simultaneously. When multiple microphones are available, spatial features can be used to improve the localization and separation accuracy of sounds from different directions; however, conventional methods have three drawbacks: (a) Sound source localization and sound source separation methods using spatial features and classification using spectral features trained in the same neural network, may overfit to the relationship between the direction of arrival and the class of a sound, thereby reducing their reliability to deal with novel events. (b) Although permutation invariant training used in autonomous speech recognition could be extended, it is impractical for environmental sounds that include an unlimited number of sound sources. (c) Various features, such as complex values of short time Fourier transform and interchannel phase differences have been used as spatial features, but no study has compared them. This paper proposes a multichannel environmental sound segmentation method comprising two discrete blocks, a sound source localization and separation block and a sound source separation and classification block. By separating the blocks, overfitting to the relationship between the direction of arrival and the class is avoided. Simulation experiments using created datasets including 75-class environmental sounds showed the root mean squared error of the proposed method was lower than that of conventional methods.

Streaming End-to-End Multi-Talker Speech Recognition

End-to-end multi-talker speech recognition is an emerging research trend in the speech community due to its vast potential in applications such as conversation and meeting transcriptions. To the best of our knowledge, all existing research works are constrained in the offline scenario. In this work, we propose the Streaming Unmixing and Recognition Transducer (SURT) for end-to-end multi-talker speech recognition. Our model employs the Recurrent Neural Network Transducer (RNN-T) as the backbone that can meet various latency constraints. We study two different model architectures that are based on a speaker-differentiator encoder and a mask encoder respectively. To train this model, we investigate the widely used Permutation Invariant Training (PIT) approach and the Heuristic Error Assignment Training (HEAT) approach. Based on experiments on the publicly available LibriSpeechMix dataset, we show that HEAT can achieve better accuracy compared with PIT, and the SURT model with 150 milliseconds algorithmic latency constraint compares favorably with the offline sequence-to-sequence based baseline model in terms of accuracy.

Speaker Separation Using Speaker Inventories and Estimated Speech

We propose speaker separation using speaker inventories and estimated speech (SSUSIES), a framework leveraging speaker profiles and estimated speech for speaker separation. SSUSIES contains two methods, speaker separation using speaker inventories (SSUSI) and speaker separation using estimated speech (SSUES). SSUSI performs speaker separation with the help of speaker inventory. By combining the advantages of permutation invariant training (PIT) and speech extraction, SSUSI significantly outperforms conventional approaches. SSUES is a widely applicable technique that can substantially improve speaker separation performance using the output of first-pass separation. We evaluate the models on both speaker separation and speech recognition metrics.

Incorporating Phase-Encoded Spectrum Masking into Speaker-Independent Monaural Source Separation

Abstract Typical speech separation systems usually operate in the time-frequency (T-F) domain by enhancing the magnitude response and leaving the phase response unaltered. Recent studies, however, suggest that phase is important for perceptual quality, leading many researchers to consider magnitude and phase spectrum enhancements. The complex-valued Fourier spectrum and real-valued shifted real spectrum (SRS) of a discrete-time speech signal sequence carry information both in magnitude and phase functions, which can reconstruct the speech signal without loss. In this paper, we propose novel methods called the pcIRM and pSRSM to solve the problem of speaker-independent monaural source separation and to recover the phase of source speech signal. Specifically, the pcIRM achieves the complex ideal ratio mask (cIRM) estimation based on the complex-valued phase-encoded Fourier spectrogram, and the pSRSM obtains the shifted real spectrum mask (SRSM) estimation based on the real-valued phase-encoded shifted real spectrum, both of which are creatively implemented with the deep bi-directional Long Short-Term Memory networks. Furthermore, the merging of these masking estimation methods and the utterance-level permutation invariant training (uPIT) approach has been proved to be an effective way to improve speech separation. Typically, the uPIT addresses the label ambiguity problem well, which is the major barrier for speaker-independent multi-talker source separation. Finally, we report separation results for the proposed methods and compare them with that of the state-of-the-art methods, which are evaluated with the WSJ0-2mix datasets. Extensive experiment results demonstrate the advantages of our proposed pcIRM method in terms of the signal-to-distortion ratio (SDR) and the perceptual evaluation of speech quality (PESQ) metrics, and the pSRSM method obtains comparable performance to the pcIRM in the opposite gender speakers source separation circumstance with smaller model complexity.

Fact-Aware Sentence Split and Rephrase with Permutation Invariant Training

Sentence Split and Rephrase aims to break down a complex sentence into several simple sentences with its meaning preserved. Previous studies tend to address the issue by seq2seq learning from parallel sentence pairs, which takes a complex sentence as input and sequentially generates a series of simple sentences. However, the conventional seq2seq learning has two limitations for this task: (1) it does not take into account the facts stated in the long sentence; As a result, the generated simple sentences may miss or inaccurately state the facts in the original sentence. (2) The order variance of the simple sentences to be generated may confuse the seq2seq model during training because the simple sentences derived from the long source sentence could be in any order.To overcome the challenges, we first propose the Fact-aware Sentence Encoding, which enables the model to learn facts from the long sentence and thus improves the precision of sentence split; then we introduce Permutation Invariant Training to alleviate the effects of order variance in seq2seq learning for this task. Experiments on the WebSplit-v1.0 benchmark dataset show that our approaches can largely improve the performance over the previous seq2seq learning approaches. Moreover, an extrinsic evaluation on oie-benchmark verifies the effectiveness of our approaches by an observation that splitting long sentences with our state-of-the-art model as preprocessing is helpful for improving OpenIE performance.

Improving End-to-End Single-Channel Multi-Talker Speech Recognition

Although significant progress has been made in single-talker automatic speech recognition (ASR), there is still a large performance gap between multi-talker and single-talker speech recognition systems. In this article, we propose an enhanced end-to-end monaural multi-talker ASR architecture and training strategy to recognize the overlapped speech. The single-talker end-to-end model is extended to a multi-talker architecture with permutation invariant training (PIT). Several methods are designed to enhance the system performance, including speaker parallel attention, scheduled sampling, curriculum learning and knowledge distillation. More specifically, the speaker parallel attention extends the basic single shared attention module into multiple attention modules for each speaker, which can enhance the tracing and separation ability. Then the scheduled sampling and curriculum learning are proposed to make the model better optimized. Finally the knowledge distillation transfers the knowledge from an original single-speaker model to the current multi-speaker model in the proposed end-to-end multi-talker ASR structure. Our proposed architectures are evaluated and compared on the artificially mixed speech datasets generated from the WSJ0 reading corpus. The experiments demonstrate that our proposed architectures can significantly improve the multi-talker mixed speech recognition. The final system obtains more than 15% relative performance gains in both character error rate (CER) and word error rate (WER) compared to the basic end-to-end multi-talker ASR system.

Audio–Visual Deep Clustering for Speech Separation

Speech separation aims to separate individual voices from an audio mixture of multiple simultaneous talkers. Audio-only approaches show unsatisfactory performance when the speakers are of the same gender or share similar voice characteristics. This is due to challenges on learning appropriate feature representations for separating voices in single frames and streaming voices across time. Visual signals of speech (e.g., lip movements), if available, can be leveraged to learn better feature representations for separation. In this paper, we propose a novel audio–visual deep clustering model (AVDC) to integrate visual information into the process of learning better feature representations (embeddings) for Time–Frequency (T–F) bin clustering. It employs a two-stage audio–visual fusion strategy where speaker-wise audio–visual T–F embeddings are first computed after the first-stage fusion to model the audio–visual correspondence for each speaker. In the second-stage fusion, audio–visual embeddings of all speakers and audio embeddings calculated by deep clustering from the audio mixture are concatenated to form the final T–F embedding for clustering. Through a series of experiments, the proposed AVDC model is shown to outperform the audio-only deep clustering and utterance-level permutation invariant training baselines and three other state-of-the-art audio–visual approaches. Further analyses show that the AVDC model learns a better T–F embedding for alleviating the source permutation problem across frames. Other experiments show that the AVDC model is able to generalize across different numbers of speakers between training and testing and shows some robustness when visual information is partially missing.

Combining Spectral and Spatial Features for Deep Learning Based Blind Speaker Separation

This study tightly integrates complementary spectral and spatial features for deep learning based multi-channel speaker separation in reverberant environments. The key idea is to localize individual speakers so that an enhancement network can be trained on spatial as well as spectral features to extract the speaker from an estimated direction and with specific spectral structures. The spatial and spectral features are designed in a way such that the trained models are blind to the number of microphones and microphone geometry. To determine the direction of the speaker of interest, we identify time-frequency T-F units dominated by that speaker and only use them for direction estimation. The T-F unit level speaker dominance is determined by a two-channel chimera++ network, which combines deep clustering and permutation invariant training at the objective function level, and integrates spectral and interchannel phase patterns at the input feature level. In addition, T-F masking based beamforming is tightly integrated in the system by leveraging the magnitudes and phases produced by beamforming. Strong separation performance has been observed on reverberant talker-independent speaker separation, which separates reverberant speaker mixtures based on a random number of microphones arranged in arbitrary linear-array geometry.

Single-channel multi-talker speech recognition with permutation invariant training

Although great progress has been made in automatic speech recognition (ASR), significant performance degradation is still observed when recognizing multi-talker mixed speech. In this paper, we propose and evaluate several architectures to address this problem under the assumption that only a single channel of mixed signal is available. Our technique extends permutation invariant training (PIT) by introducing the front-end feature separation module with the minimum mean square error (MSE) criterion and the back-end recognition module with the minimum cross entropy (CE) criterion. More specifically, during training we compute the average MSE or CE over the whole utterance for each possible utterance-level output-target assignment, pick the one with the minimum MSE or CE, and optimize for that assignment. This strategy elegantly solves the label permutation problem observed in the deep learning based multi-talker mixed speech separation and recognition systems. The proposed architectures are evaluated and compared on an artificially mixed AMI dataset with both two- and three-talker mixed speech. The experimental results indicate that against the state-of-the-art single-talker speech recognition system our proposed architectures can cut the word error rate (WER) by relative 45.0% and 25.0% across all speakers when their energies are comparable, for two- and three-talker mixed speech, respectively. To our knowledge, this is the first work on the single-channel multi-talker mixed speech recognition on the challenging speaker-independent spontaneous large vocabulary continuous speech task.

Progressive Joint Modeling in Unsupervised Single-Channel Overlapped Speech Recognition

Unsupervised single-channel overlapped speech recognition is one of the hardest problems in automatic speech recognition (ASR). Permutation invariant training (PIT) is a state of the art model-based approach, which applies a single neural network to solve this single-input, multiple-output modeling problem. We propose to advance the current state of the art by imposing a modular structure on the neural network, applying a progressive pretraining regimen, and improving the objective function with transfer learning and a discriminative training criterion. The modular structure splits the problem into three sub-tasks: frame-wise interpreting, utterance-level speaker tracing, and speech recognition. The pretraining regimen uses these modules to solve progressively harder tasks. Transfer learning leverages parallel clean speech to improve the training targets for the network. Our discriminative training formulation is a modification of standard formulations, that also penalizes competing outputs of the system. Experiments are conducted on the artificial overlapped Switchboard and hub5e-swb dataset. The proposed framework achieves over 30% relative improvement of WER over both a strong jointly trained system, PIT for ASR, and a separately optimized system, PIT for speech separation with clean speech ASR model. The improvement comes from better model generalization, training efficiency and the sequence level linguistic knowledge integration.

Past review, current progress, and challenges ahead on the cocktail party problem

The cocktail party problem, i.e., tracing and recognizing the speech of a specific speaker when multiple speakers talk simultaneously, is one of the critical problems yet to be solved to enable the wide application of automatic speech recognition (ASR) systems. In this overview paper, we review the techniques proposed in the last two decades in attacking this problem. We focus our discussions on the speech separation problem given its central role in the cocktail party environment, and describe the conventional single-channel techniques such as computational auditory scene analysis (CASA), non-negative matrix factorization (NMF) and generative models, the conventional multi-channel techniques such as beamforming and multi-channel blind source separation, and the newly developed deep learning-based techniques, such as deep clustering (DPCL), the deep attractor network (DANet), and permutation invariant training (PIT). We also present techniques developed to improve ASR accuracy and speaker identification in the cocktail party environment. We argue effectively exploiting information in the microphone array, the acoustic training set, and the language itself using a more powerful model. Better optimization objective and techniques will be the approach to solving the cocktail party problem.

Multitalker Speech Separation With Utterance-Level Permutation Invariant Training of Deep Recurrent Neural Networks

In this paper, we propose the utterance-level permutation invariant training (uPIT) technique. uPIT is a practically applicable, end-to-end, deep-learning-based solution for speaker independent multitalker speech separation. Specifically, uPIT extends the recently proposed permutation invariant training (PIT) technique with an utterance-level cost function, hence eliminating the need for solving an additional permutation problem during inference, which is otherwise required by frame-level PIT. We achieve this using recurrent neural networks (RNNs) that, during training, minimize the utterance-level separation error, hence forcing separated frames belonging to the same speaker to be aligned to the same output stream. In practice, this allows RNNs, trained with uPIT, to separate multitalker mixed speech without any prior knowledge of signal duration, number of speakers, speaker identity, or gender. We evaluated uPIT on the WSJ0 and Danish two- and three-talker mixed-speech separation tasks and found that uPIT outperforms techniques based on nonnegative matrix factorization and computational auditory scene analysis, and compares favorably with deep clustering, and the deep attractor network. Furthermore, we found that models trained with uPIT generalize well to unseen speakers and languages. Finally, we found that a single model, trained with uPIT, can handle both two-speaker, and three-speaker speech mixtures.