Reinhold Haeb-Umbach

Thilo von Neumann, Christoph Boeddeker, Marc Delcroix et al.

MeetEval is an open-source toolkit to evaluate all kinds of meeting transcription systems. It provides a unified interface for the computation of commonly used Word Error Rates (WERs), specifically cpWER, ORC-WER and MIMO-WER along other WER definitions. We extend the cpWER computation by a temporal constraint to ensure that only words are identified as correct when the temporal alignment is plausible. This leads to a better quality of the matching of the hypothesis string to the reference string that more closely resembles the actual transcription quality, and a system is penalized if it provides poor time annotations. Since word-level timing information is often not available, we present a way to approximate exact word-level timings from segment-level timings (e.g., a sentence) and show that the approximation leads to a similar WER as a matching with exact word-level annotations. At the same time, the time constraint leads to a speedup of the matching algorithm, which outweighs the additional overhead caused by processing the time stamps.

Christoph Boeddeker, Patrick Hanebrink, Lukas Drude et al.

This report describes the computation of gradients by algorithmic differentiation for statistically optimum beamforming operations. Especially the derivation of complex-valued functions is a key component of this approach. Therefore the real-valued algorithmic differentiation is extended via the complex-valued chain rule. In addition to the basic mathematic operations the derivative of the eigenvalue problem with complex-valued eigenvectors is one of the key results of this report. The potential of this approach is shown with experimental results on the CHiME-3 challenge database. There, the beamforming task is used as a front-end for an ASR system. With the developed derivatives a joint optimization of a speech enhancement and speech recognition system w.r.t. the recognition optimization criterion is possible.

Simon Berger, Peter Vieting, Christoph Boeddeker et al.

Multi-speaker automatic speech recognition (ASR) is crucial for many real-world applications, but it requires dedicated modeling techniques. Existing approaches can be divided into modular and end-to-end methods. Modular approaches separate speakers and recognize each of them with a single-speaker ASR system. End-to-end models process overlapped speech directly in a single, powerful neural network. This work proposes a middle-ground approach that leverages explicit speech separation similarly to the modular approach but also incorporates mixture speech information directly into the ASR module in order to mitigate the propagation of errors made by the speech separator. We also explore a way to exchange cross-speaker context information through a layer that combines information of the individual speakers. Our system is optimized through separate and joint training stages and achieves a relative improvement of 7% in word error rate over a purely modular setup on the SMS-WSJ task.

Peter Vieting, Simon Berger, Thilo von Neumann et al.

Many real-life applications of automatic speech recognition (ASR) require processing of overlapped speech. A common method involves first separating the speech into overlap-free streams on which ASR is performed. Recently, TF-GridNet has shown impressive performance in speech separation in real reverberant conditions. Furthermore, a mixture encoder was proposed that leverages the mixed speech to mitigate the effect of separation artifacts. In this work, we extended the mixture encoder from a static two-speaker scenario to a natural meeting context featuring an arbitrary number of speakers and varying degrees of overlap. We further demonstrate its limits by the integration with separators of varying strength including TF-GridNet. Our experiments result in a new state-of-the-art performance on LibriCSS using a single microphone. They show that TF-GridNet largely closes the gap between previous methods and oracle separation independent of mixture encoding. We further investigate the remaining potential for improvement.

Joerg Schmalenstroeer, Tobias Gburrek, Reinhold Haeb-Umbach

We present LibriWASN, a data set whose design follows closely the LibriCSS meeting recognition data set, with the marked difference that the data is recorded with devices that are randomly positioned on a meeting table and whose sampling clocks are not synchronized. Nine different devices, five smartphones with a single recording channel and four microphone arrays, are used to record a total of 29 channels. Other than that, the data set follows closely the LibriCSS design: the same LibriSpeech sentences are played back from eight loudspeakers arranged around a meeting table and the data is organized in subsets with different percentages of speech overlap. LibriWASN is meant as a test set for clock synchronization algorithms, meeting separation, diarization and transcription systems on ad-hoc wireless acoustic sensor networks. Due to its similarity to LibriCSS, meeting transcription systems developed for the former can readily be tested on LibriWASN. The data set is recorded in two different rooms and is complemented with ground-truth diarization information of who speaks when.

Janek Ebbers, Reinhold Haeb-Umbach

In this paper we present our system for the detection and classification of acoustic scenes and events (DCASE) 2020 Challenge Task 4: Sound event detection and separation in domestic environments. We introduce two new models: the forward-backward convolutional recurrent neural network (FBCRNN) and the tag-conditioned convolutional neural network (CNN). The FBCRNN employs two recurrent neural network (RNN) classifiers sharing the same CNN for preprocessing. With one RNN processing a recording in forward direction and the other in backward direction, the two networks are trained to jointly predict audio tags, i.e., weak labels, at each time step within a recording, given that at each time step they have jointly processed the whole recording. The proposed training encourages the classifiers to tag events as soon as possible. Therefore, after training, the networks can be applied to shorter audio segments of, e.g., 200 ms, allowing sound event detection (SED). Further, we propose a tag-conditioned CNN to complement SED. It is trained to predict strong labels while using (predicted) tags, i.e., weak labels, as additional input. For training pseudo strong labels from a FBCRNN ensemble are used. The presented system scored the fourth and third place in the systems and teams rankings, respectively. Subsequent improvements allow our system to even outperform the challenge baseline and winner systems in average by, respectively, 18.0% and 2.2% event-based F1-score on the validation set. Source code is publicly available at https://github.com/fgnt/pb_sed.

Peter Vieting, Simon Berger, Thilo von Neumann et al.

Meeting transcription is a field of high relevance and remarkable progress in recent years. Still, challenges remain that limit its performance. In this work, we extend a previously proposed framework for analyzing leakage in speech separation with proper sensitivity to temporal locality. We show that there is significant leakage to the cross channel in areas where only the primary speaker is active. At the same time, the results demonstrate that this does not affect the final performance much as these leaked parts are largely ignored by the voice activity detection (VAD). Furthermore, different segmentations are compared showing that advanced diarization approaches are able to reduce the gap to oracle segmentation by a third compared to a simple energy-based VAD. We additionally reveal what factors contribute to the remaining difference. The results represent state-of-the-art performance on LibriCSS among systems that train the recognition module on LibriSpeech data only.

Janek Ebbers, Romain Serizel, Reinhold Haeb-Umbach

Performing an adequate evaluation of sound event detection (SED) systems is far from trivial and is still subject to ongoing research. The recently proposed polyphonic sound detection (PSD)-receiver operating characteristic (ROC) and PSD score (PSDS) make an important step into the direction of an evaluation of SED systems which is independent from a certain decision threshold. This allows to obtain a more complete picture of the overall system behavior which is less biased by threshold tuning. Yet, the PSD-ROC is currently only approximated using a finite set of thresholds. The choice of the thresholds used in approximation, however, can have a severe impact on the resulting PSDS. In this paper we propose a method which allows for computing system performance on an evaluation set for all possible thresholds jointly, enabling accurate computation not only of the PSD-ROC and PSDS but also of other collar-based and intersection-based performance curves. It further allows to select the threshold which best fulfills the requirements of a given application. Source code is publicly available in our SED evaluation package sed_scores_eval.

Tobias Cord-Landwehr, Christoph Boeddeker, Thilo von Neumann et al.

Impressive progress in neural network-based single-channel speech source separation has been made in recent years. But those improvements have been mostly reported on anechoic data, a situation that is hardly met in practice. Taking the SepFormer as a starting point, which achieves state-of-the-art performance on anechoic mixtures, we gradually modify it to optimize its performance on reverberant mixtures. Although this leads to a word error rate improvement by 7 percentage points compared to the standard SepFormer implementation, the system ends up with only marginally better performance than a PIT-BLSTM separation system, that is optimized with rather straightforward means. This is surprising and at the same time sobering, challenging the practical usefulness of many improvements reported in recent years for monaural source separation on nonreverberant data.

Thilo von Neumann, Keisuke Kinoshita, Christoph Boeddeker et al.

Many state-of-the-art neural network-based source separation systems use the averaged Signal-to-Distortion Ratio (SDR) as a training objective function. The basic SDR is, however, undefined if the network reconstructs the reference signal perfectly or if the reference signal contains silence, e.g., when a two-output separator processes a single-speaker recording. Many modifications to the plain SDR have been proposed that trade-off between making the loss more robust and distorting its value. We propose to switch from a mean over the SDRs of each individual output channel to a global SDR over all output channels at the same time, which we call source-aggregated SDR (SA-SDR). This makes the loss robust against silence and perfect reconstruction as long as at least one reference signal is not silent. We experimentally show that our proposed SA-SDR is more stable and preferable over other well-known modifications when processing meeting-style data that typically contains many silent or single-speaker regions.

Tobias Gburrek, Joerg Schmalenstroeer, Reinhold Haeb-Umbach

A wireless acoustic sensor network records audio signals with sampling time and sampling rate offsets between the audio streams, if the analog-digital converters (ADCs) of the network devices are not synchronized. Here, we introduce a new sampling rate offset model to simulate time-varying sampling frequencies caused, for example, by temperature changes of ADC crystal oscillators, and propose an estimation algorithm to handle this dynamic aspect in combination with changing acoustic source positions. Furthermore, we show how deep neural network based estimates of the distances between microphones and human speakers can be used to determine the sampling time offsets. This enables a synchronization of the audio streams to reflect the physical time differences of flight.

Thilo von Neumann, Keisuke Kinoshita, Christoph Boeddeker et al.

Automatic transcription of meetings requires handling of overlapped speech, which calls for continuous speech separation (CSS) systems. The uPIT criterion was proposed for utterance-level separation with neural networks and introduces the constraint that the total number of speakers must not exceed the number of output channels. When processing meeting-like data in a segment-wise manner, i.e., by separating overlapping segments independently and stitching adjacent segments to continuous output streams, this constraint has to be fulfilled for any segment. In this contribution, we show that this constraint can be significantly relaxed. We propose a novel graph-based PIT criterion, which casts the assignment of utterances to output channels in a graph coloring problem. It only requires that the number of concurrently active speakers must not exceed the number of output channels. As a consequence, the system can process an arbitrary number of speakers and arbitrarily long segments and thus can handle more diverse scenarios. Further, the stitching algorithm for obtaining a consistent output order in neighboring segments is of less importance and can even be eliminated completely, not the least reducing the computational effort. Experiments on meeting-style WSJ data show improvements in recognition performance over using the uPIT criterion.

Thilo von Neumann, Christoph Boeddeker, Keisuke Kinoshita et al.

Permutation invariant training (PIT) is a widely used training criterion for neural network-based source separation, used for both utterance-level separation with utterance-level PIT (uPIT) and separation of long recordings with the recently proposed Graph-PIT. When implemented naively, both suffer from an exponential complexity in the number of utterances to separate, rendering them unusable for large numbers of speakers or long realistic recordings. We present a decomposition of the PIT criterion into the computation of a matrix and a strictly monotonously increasing function so that the permutation or assignment problem can be solved efficiently with several search algorithms. The Hungarian algorithm can be used for uPIT and we introduce various algorithms for the Graph-PIT assignment problem to reduce the complexity to be polynomial in the number of utterances.

Christoph Boeddeker, Frederik Rautenberg, Reinhold Haeb-Umbach

Unsupervised blind source separation methods do not require a training phase and thus cannot suffer from a train-test mismatch, which is a common concern in neural network based source separation. The unsupervised techniques can be categorized in two classes, those building upon the sparsity of speech in the Short-Time Fourier transform domain and those exploiting non-Gaussianity or non-stationarity of the source signals. In this contribution, spatial mixture models which fall in the first category and independent vector analysis (IVA) as a representative of the second category are compared w.r.t. their separation performance and the performance of a downstream speech recognizer on a reverberant dataset of reasonable size. Furthermore, we introduce a serial concatenation of the two, where the result of the mixture model serves as initialization of IVA, which achieves significantly better WER performance than each algorithm individually and even approaches the performance of a much more complex neural network based technique.

Jens Heitkaemper, Joerg Schmalenstroeer, Joerg Ullmann et al.

In this paper we present an open database for the development of detection and enhancement algorithms of speech transmitted over HF radio channels. It consists of audio samples recorded by various receivers at different locations across Europe, all monitoring the same single-sideband modulated transmission from a base station in Paderborn, Germany. Transmitted and received speech signals are precisely time aligned to offer parallel data for supervised training of deep learning based detection and enhancement algorithms. For the task of speech activity detection two exemplary baseline systems are presented, one based on statistical methods employing a multi-stage Wiener filter with minimum statistics noise floor estimation, and the other relying on a deep learning approach.

Joerg Schmalenstroeer, Jens Heitkaemper, Joerg Ullmann et al.

In this paper we investigate the task of detecting carrier frequency differences from demodulated single sideband signals by examining the pitch contours of the received baseband speech signal in the short-time spectral domain. From the detected pitch frequency trajectory and its harmonics a carrier frequency difference, which is caused by demodulating the radio signal with the wrong carrier frequency, can be deduced. A computationally efficient realization in the power cepstral domain is presented. The core component, i.e., the pitch tracking algorithm, is shown to perform comparably to a state of the art algorithm. The full carrier frequency difference estimation system is tested on recordings of real transmissions over HF links. A comparison with an existing approach shows improved estimation accuracy, both on short and longer speech utterances

Wangyou Zhang, Christoph Boeddeker, Shinji Watanabe et al.

Recently, the end-to-end approach has been successfully applied to multi-speaker speech separation and recognition in both single-channel and multichannel conditions. However, severe performance degradation is still observed in the reverberant and noisy scenarios, and there is still a large performance gap between anechoic and reverberant conditions. In this work, we focus on the multichannel multi-speaker reverberant condition, and propose to extend our previous framework for end-to-end dereverberation, beamforming, and speech recognition with improved numerical stability and advanced frontend subnetworks including voice activity detection like masks. The techniques significantly stabilize the end-to-end training process. The experiments on the spatialized wsj1-2mix corpus show that the proposed system achieves about 35% WER relative reduction compared to our conventional multi-channel E2E ASR system, and also obtains decent speech dereverberation and separation performance (SDR=12.5 dB) in the reverberant multi-speaker condition while trained only with the ASR criterion.

Christopher Grimm, Tai Fei, Ernst Warsitz et al.

We present an approach to automatically generate semantic labels for real recordings of automotive range-Doppler (RD) radar spectra. Such labels are required when training a neural network for object recognition from radar data. The automatic labeling approach rests on the simultaneous recording of camera and lidar data in addition to the radar spectrum. By warping radar spectra into the camera image, state-of-the-art object recognition algorithms can be applied to label relevant objects, such as cars, in the camera image. The warping operation is designed to be fully differentiable, which allows backpropagating the gradient computed on the camera image through the warping operation to the neural network operating on the radar data. As the warping operation relies on accurate scene flow estimation, we further propose a novel scene flow estimation algorithm which exploits information from camera, lidar and radar sensors. The proposed scene flow estimation approach is compared against a state-of-the-art scene flow algorithm, and it outperforms it by approximately 30% w.r.t. mean average error. The feasibility of the overall framework for automatic label generation for RD spectra is verified by evaluating the performance of neural networks trained with the proposed framework for Direction-of-Arrival estimation.

Tobias Gburrek, Joerg Schmalenstroeer, Reinhold Haeb-Umbach

In this paper we present an approach to geometry calibration in wireless acoustic sensor networks, whose nodes are assumed to be equipped with a compact microphone array. The proposed approach solely works with estimates of the distances between acoustic sources and the nodes that record these sources. It consists of an iterative weighted least squares localization procedure, which is initialized by multidimensional scaling. Alongside the sensor node locations, also the positions of the acoustic sources are estimated. Furthermore, we derive the Cramer-Rao lower bound (CRLB) for source and sensor position estimation, and show by simulation that the estimator is efficient.

Christoph Boeddeker, Wangyou Zhang, Tomohiro Nakatani et al.

Time-domain training criteria have proven to be very effective for the separation of single-channel non-reverberant speech mixtures. Likewise, mask-based beamforming has shown impressive performance in multi-channel reverberant speech enhancement and source separation. Here, we propose to combine neural network supported multi-channel source separation with a time-domain training objective function. For the objective we propose to use a convolutive transfer function invariant Signal-to-Distortion Ratio (CI-SDR) based loss. While this is a well-known evaluation metric (BSS Eval), it has not been used as a training objective before. To show the effectiveness, we demonstrate the performance on LibriSpeech based reverberant mixtures. On this task, the proposed system approaches the error rate obtained on single-source non-reverberant input, i.e., LibriSpeech test_clean, with a difference of only 1.2 percentage points, thus outperforming a conventional permutation invariant training based system and alternative objectives like Scale Invariant Signal-to-Distortion Ratio by a large margin.

Tobias Gburrek, Joerg Schmalenstroeer, Andreas Brendel et al.

We present an approach to deep neural network based (DNN-based) distance estimation in reverberant rooms for supporting geometry calibration tasks in wireless acoustic sensor networks. Signal diffuseness information from acoustic signals is aggregated via the coherent-to-diffuse power ratio to obtain a distance-related feature, which is mapped to a source-to-microphone distance estimate by means of a DNN. This information is then combined with direction-of-arrival estimates from compact microphone arrays to infer the geometry of the sensor network. Unlike many other approaches to geometry calibration, the proposed scheme does only require that the sampling clocks of the sensor nodes are roughly synchronized. In simulations we show that the proposed DNN-based distance estimator generalizes to unseen acoustic environments and that precise estimates of the sensor node positions are obtained.

Keisuke Kinoshita, Thilo von Neumann, Marc Delcroix et al.

Recently, the source separation performance was greatly improved by time-domain audio source separation based on dual-path recurrent neural network (DPRNN). DPRNN is a simple but effective model for a long sequential data. While DPRNN is quite efficient in modeling a sequential data of the length of an utterance, i.e., about 5 to 10 second data, it is harder to apply it to longer sequences such as whole conversations consisting of multiple utterances. It is simply because, in such a case, the number of time steps consumed by its internal module called inter-chunk RNN becomes extremely large. To mitigate this problem, this paper proposes a multi-path RNN (MPRNN), a generalized version of DPRNN, that models the input data in a hierarchical manner. In the MPRNN framework, the input data is represented at several (>3) time-resolutions, each of which is modeled by a specific RNN sub-module. For example, the RNN sub-module that deals with the finest resolution may model temporal relationship only within a phoneme, while the RNN sub-module handling the most coarse resolution may capture only the relationship between utterances such as speaker information. We perform experiments using simulated dialogue-like mixtures and show that MPRNN has greater model capacity, and it outperforms the current state-of-the-art DPRNN framework especially in online processing scenarios.

Thilo von Neumann, Christoph Boeddeker, Lukas Drude et al.

Most approaches to multi-talker overlapped speech separation and recognition assume that the number of simultaneously active speakers is given, but in realistic situations, it is typically unknown. To cope with this, we extend an iterative speech extraction system with mechanisms to count the number of sources and combine it with a single-talker speech recognizer to form the first end-to-end multi-talker automatic speech recognition system for an unknown number of active speakers. Our experiments show very promising performance in counting accuracy, source separation and speech recognition on simulated clean mixtures from WSJ0-2mix and WSJ0-3mix. Among others, we set a new state-of-the-art word error rate on the WSJ0-2mix database. Furthermore, our system generalizes well to a larger number of speakers than it ever saw during training, as shown in experiments with the WSJ0-4mix database.

Janek Ebbers, Michael Kuhlmann, Tobias Cord-Landwehr et al.

In this work we address disentanglement of style and content in speech signals. We propose a fully convolutional variational autoencoder employing two encoders: a content encoder and a style encoder. To foster disentanglement, we propose adversarial contrastive predictive coding. This new disentanglement method does neither need parallel data nor any supervision. We show that the proposed technique is capable of separating speaker and content traits into the two different representations and show competitive speaker-content disentanglement performance compared to other unsupervised approaches. We further demonstrate an increased robustness of the content representation against a train-test mismatch compared to spectral features, when used for phone recognition.

Jens Heitkaemper, Joerg Schmalenstroeer, Reinhold Haeb-Umbach

Speech activity detection (SAD), which often rests on the fact that the noise is "more" stationary than speech, is particularly challenging in non-stationary environments, because the time variance of the acoustic scene makes it difficult to discriminate speech from noise. We propose two approaches to SAD, where one is based on statistical signal processing, while the other utilizes neural networks. The former employes sophisticated signal processing to track the noise and speech energies and is meant to support the case for a resource efficient, unsupervised signal processing approach. The latter introduces a recurrent network layer that operates on short segments of the input speech to do temporal smoothing in the presence of non-stationary noise. The systems are tested on the Fearless Steps challenge, which consists of the transmission data from the Apollo-11 space mission. The statistical SAD achieves comparable detection performance to earlier proposed neural network based SADs, while the neural network based approach leads to a decision cost function of 1.07% on the evaluation set of the 2020 Fearless Steps Challenge, which sets a new state of the art.

Tomohiro Nakatani, Christoph Boeddeker, Keisuke Kinoshita et al.

This paper proposes methods that can optimize a Convolutional BeamFormer (CBF) for jointly performing denoising, dereverberation, and source separation (DN+DR+SS) in a computationally efficient way. Conventionally, cascade configuration composed of a Weighted Prediction Error minimization (WPE) dereverberation filter followed by a Minimum Variance Distortionless Response beamformer has been usedas the state-of-the-art frontend of far-field speech recognition, however, overall optimality of this approach is not guaranteed. In the blind signal processing area, an approach for jointly optimizing dereverberation and source separation (DR+SS) has been proposed, however, this approach requires huge computing cost, and has not been extended for application to DN+DR+SS. To overcome the above limitations, this paper develops new approaches for jointly optimizing DN+DR+SS in a computationally much more efficient way. To this end, we first present an objective function to optimize a CBF for performing DN+DR+SS based on the maximum likelihood estimation, on an assumption that the steering vectors of the target signals are given or can be estimated, e.g., using a neural network. This paper refers to a CBF optimized by this objective function as a weighted Minimum-Power Distortionless Response (wMPDR) CBF. Then, we derive two algorithms for optimizing a wMPDR CBF based on two different ways of factorizing a CBF into WPE filters and beamformers. Experiments using noisy reverberant sound mixtures show that the proposed optimization approaches greatly improve the performance of the speech enhancement in comparison with the conventional cascade configuration in terms of the signal distortion measures and ASR performance. It is also shown that the proposed approaches can greatly reduce the computing cost with improved estimation accuracy in comparison with the conventional joint optimization approach.

Thilo von Neumann, Keisuke Kinoshita, Lukas Drude et al.

The rising interest in single-channel multi-speaker speech separation sparked development of End-to-End (E2E) approaches to multi-speaker speech recognition. However, up until now, state-of-the-art neural network-based time domain source separation has not yet been combined with E2E speech recognition. We here demonstrate how to combine a separation module based on a Convolutional Time domain Audio Separation Network (Conv-TasNet) with an E2E speech recognizer and how to train such a model jointly by distributing it over multiple GPUs or by approximating truncated back-propagation for the convolutional front-end. To put this work into perspective and illustrate the complexity of the design space, we provide a compact overview of single-channel multi-speaker recognition systems. Our experiments show a word error rate of 11.0% on WSJ0-2mix and indicate that our joint time domain model can yield substantial improvements over cascade DNN-HMM and monolithic E2E frequency domain systems proposed so far.

Jens Heitkaemper, Darius Jakobeit, Christoph Boeddeker et al.

In recent years time domain speech separation has excelled over frequency domain separation in single channel scenarios and noise-free environments. In this paper we dissect the gains of the time-domain audio separation network (TasNet) approach by gradually replacing components of an utterance-level permutation invariant training (u-PIT) based separation system in the frequency domain until the TasNet system is reached, thus blending components of frequency domain approaches with those of time domain approaches. Some of the intermediate variants achieve comparable signal-to-distortion ratio (SDR) gains to TasNet, but retain the advantage of frequency domain processing: compatibility with classic signal processing tools such as frequency-domain beamforming and the human interpretability of the masks. Furthermore, we show that the scale invariant signal-to-distortion ratio (si-SDR) criterion used as loss function in TasNet is related to a logarithmic mean square error criterion and that it is this criterion which contributes most reliable to the performance advantage of TasNet. Finally, we critically assess which gains in a noise-free single channel environment generalize to more realistic reverberant conditions.

Lukas Drude, Jens Heitkaemper, Christoph Boeddeker et al.

We present a multi-channel database of overlapping speech for training, evaluation, and detailed analysis of source separation and extraction algorithms: SMS-WSJ -- Spatialized Multi-Speaker Wall Street Journal. It consists of artificially mixed speech taken from the WSJ database, but unlike earlier databases we consider all WSJ0+1 utterances and take care of strictly separating the speaker sets present in the training, validation and test sets. When spatializing the data we ensure a high degree of randomness w.r.t. room size, array center and rotation, as well as speaker position. Furthermore, this paper offers a critical assessment of recently proposed measures of source separation performance. Alongside the code to generate the database we provide a source separation baseline and a Kaldi recipe with competitive word error rates to provide common ground for evaluation.

Christoph Boeddeker, Tomohiro Nakatani, Keisuke Kinoshita et al.

We previously proposed an optimal (in the maximum likelihood sense) convolutional beamformer that can perform simultaneous denoising and dereverberation, and showed its superiority over the widely used cascade of a WPE dereverberation filter and a conventional MPDR beamformer. However, it has not been fully investigated which components in the convolutional beamformer yield such superiority. To this end, this paper presents a new derivation of the convolutional beamformer that allows us to factorize it into a WPE dereverberation filter, and a special type of a (non-convolutional) beamformer, referred to as a wMPDR beamformer, without loss of optimality. With experiments, we show that the superiority of the convolutional beamformer in fact comes from its wMPDR part.

Catalin Zorila, Christoph Boeddeker, Rama Doddipatla et al.

Despite the strong modeling power of neural network acoustic models, speech enhancement has been shown to deliver additional word error rate improvements if multi-channel data is available. However, there has been a longstanding debate whether enhancement should also be carried out on the ASR training data. In an extensive experimental evaluation on the acoustically very challenging CHiME-5 dinner party data we show that: (i) cleaning up the training data can lead to substantial error rate reductions, and (ii) enhancement in training is advisable as long as enhancement in test is at least as strong as in training. This approach stands in contrast and delivers larger gains than the common strategy reported in the literature to augment the training database with additional artificially degraded speech. Together with an acoustic model topology consisting of initial CNN layers followed by factorized TDNN layers we achieve with 41.6% and 43.2% WER on the DEV and EVAL test sets, respectively, a new single-system state-of-the-art result on the CHiME-5 data. This is a 8% relative improvement compared to the best word error rate published so far for a speech recognizer without system combination.

Naoyuki Kanda, Christoph Boeddeker, Jens Heitkaemper et al.

In this paper, we present Hitachi and Paderborn University's joint effort for automatic speech recognition (ASR) in a dinner party scenario. The main challenges of ASR systems for dinner party recordings obtained by multiple microphone arrays are (1) heavy speech overlaps, (2) severe noise and reverberation, (3) very natural conversational content, and possibly (4) insufficient training data. As an example of a dinner party scenario, we have chosen the data presented during the CHiME-5 speech recognition challenge, where the baseline ASR had a 73.3% word error rate (WER), and even the best performing system at the CHiME-5 challenge had a 46.1% WER. We extensively investigated a combination of the guided source separation-based speech enhancement technique and an already proposed strong ASR backend and found that a tight combination of these techniques provided substantial accuracy improvements. Our final system achieved WERs of 39.94% and 41.64% for the development and evaluation data, respectively, both of which are the best published results for the dataset. We also investigated with additional training data on the official small data in the CHiME-5 corpus to assess the intrinsic difficulty of this ASR task.

Lukas Drude, Jahn Heymann, Reinhold Haeb-Umbach

We present an unsupervised training approach for a neural network-based mask estimator in an acoustic beamforming application. The network is trained to maximize a likelihood criterion derived from a spatial mixture model of the observations. It is trained from scratch without requiring any parallel data consisting of degraded input and clean training targets. Thus, training can be carried out on real recordings of noisy speech rather than simulated ones. In contrast to previous work on unsupervised training of neural mask estimators, our approach avoids the need for a possibly pre-trained teacher model entirely. We demonstrate the effectiveness of our approach by speech recognition experiments on two different datasets: one mainly deteriorated by noise (CHiME 4) and one by reverberation (REVERB). The results show that the performance of the proposed system is on par with a supervised system using oracle target masks for training and with a system trained using a model-based teacher.

Lukas Drude, Daniel Hasenklever, Reinhold Haeb-Umbach

We propose a training scheme to train neural network-based source separation algorithms from scratch when parallel clean data is unavailable. In particular, we demonstrate that an unsupervised spatial clustering algorithm is sufficient to guide the training of a deep clustering system. We argue that previous work on deep clustering requires strong supervision and elaborate on why this is a limitation. We demonstrate that (a) the single-channel deep clustering system trained according to the proposed scheme alone is able to achieve a similar performance as the multi-channel teacher in terms of word error rates and (b) initializing the spatial clustering approach with the deep clustering result yields a relative word error rate reduction of 26 % over the unsupervised teacher.

Thilo von Neumann, Keisuke Kinoshita, Marc Delcroix et al.

Automatic meeting analysis comprises the tasks of speaker counting, speaker diarization, and the separation of overlapped speech, followed by automatic speech recognition. This all has to be carried out on arbitrarily long sessions and, ideally, in an online or block-online manner. While significant progress has been made on individual tasks, this paper presents for the first time an all-neural approach to simultaneous speaker counting, diarization and source separation. The NN-based estimator operates in a block-online fashion and tracks speakers even if they remain silent for a number of time blocks, thus learning a stable output order for the separated sources. The neural network is recurrent over time as well as over the number of sources. The simulation experiments show that state of the art separation performance is achieved, while at the same time delivering good diarization and source counting results. It even generalizes well to an unseen large number of blocks.

Gerhard Kurz, Igor Gilitschenski, Florian Pfaff et al.

In this paper, we present libDirectional, a MATLAB library for directional statistics and directional estimation. It supports a variety of commonly used distributions on the unit circle, such as the von Mises, wrapped normal, and wrapped Cauchy distributions. Furthermore, various distributions on higher-dimensional manifolds such as the unit hypersphere and the hypertorus are available. Based on these distributions, several recursive filtering algorithms in libDirectional allow estimation on these manifolds. The functionality is implemented in a clear, well-documented, and object-oriented structure that is both easy to use and easy to extend.

Florian Jacob, Reinhold Haeb-Umbach

Joint audio-visual speaker tracking requires that the locations of microphones and cameras are known and that they are given in a common coordinate system. Sensor self-localization algorithms, however, are usually separately developed for either the acoustic or the visual modality and return their positions in a modality specific coordinate system, often with an unknown rotation, scaling and translation between the two. In this paper we propose two techniques to determine the positions of acoustic sensors in a common coordinate system, based on audio-visual correlates, i.e., events that are localized by both, microphones and cameras separately. The first approach maps the output of an acoustic self-calibration algorithm by estimating rotation, scale and translation to the visual coordinate system, while the second solves a joint system of equations with acoustic and visual directions of arrival as input. The evaluation of the two strategies reveals that joint calibration outperforms the mapping approach and achieves an overall calibration error of 0.20m even in reverberant environments.