Tanel Pärnamaa

Evgenii Indenbom, Nicolae-Catalin Ristea, Ando Saabas et al.

Acoustic echo cancellation (AEC), noise suppression (NS) and dereverberation (DR) are an integral part of modern full-duplex communication systems. As the demand for teleconferencing systems increases, addressing these tasks is required for an effective and efficient online meeting experience. Most prior research proposes solutions for these tasks separately, combining them with digital signal processing (DSP) based components, resulting in complex pipelines that are often impractical to deploy in real-world applications. This paper proposes a real-time cross-attention deep model, named DeepVQE, based on residual convolutional neural networks (CNNs) and recurrent neural networks (RNNs) to simultaneously address AEC, NS, and DR. We conduct several ablation studies to analyze the contributions of different components of our model to the overall performance. DeepVQE achieves state-of-the-art performance on non-personalized tracks from the ICASSP 2023 Acoustic Echo Cancellation Challenge and ICASSP 2023 Deep Noise Suppression Challenge test sets, showing that a single model can handle multiple tasks with excellent performance. Moreover, the model runs in real-time and has been successfully tested for the Microsoft Teams platform.

Evgenii Indenbom, Nicolae-Cătălin Ristea, Ando Saabas et al.

With recent research advances, deep learning models have become an attractive choice for acoustic echo cancellation (AEC) in real-time teleconferencing applications. Since acoustic echo is one of the major sources of poor audio quality, a wide variety of deep models have been proposed. However, an important but often omitted requirement for good echo cancellation quality is the synchronization of the microphone and far end signals. Typically implemented using classical algorithms based on cross-correlation, the alignment module is a separate functional block with known design limitations. In our work we propose a deep learning architecture with built-in self-attention based alignment, which is able to handle unaligned inputs, improving echo cancellation performance while simplifying the communication pipeline. Moreover, we show that our approach achieves significant improvements for difficult delay estimation cases on real recordings from AEC Challenge data set.

Ross Cutler, Ando Saabas, Tanel Parnamaa et al.

The ICASSP 2022 Acoustic Echo Cancellation Challenge is intended to stimulate research in acoustic echo cancellation (AEC), which is an important area of speech enhancement and still a top issue in audio communication. This is the third AEC challenge and it is enhanced by including mobile scenarios, adding speech recognition rate in the challenge goal metrics, and making the default sample rate 48 kHz. In this challenge, we open source two large datasets to train AEC models under both single talk and double talk scenarios. These datasets consist of recordings from more than 10,000 real audio devices and human speakers in real environments, as well as a synthetic dataset. We also open source an online subjective test framework and provide an online objective metric service for researchers to quickly test their results. The winners of this challenge are selected based on the average Mean Opinion Score achieved across all different single talk and double talk scenarios, and the speech recognition word acceptance rate.

Kusha Sridhar, Ross Cutler, Ando Saabas et al.

The ICASSP 2021 Acoustic Echo Cancellation Challenge is intended to stimulate research in the area of acoustic echo cancellation (AEC), which is an important part of speech enhancement and still a top issue in audio communication and conferencing systems. Many recent AEC studies report good performance on synthetic datasets where the train and test samples come from the same underlying distribution. However, the AEC performance often degrades significantly on real recordings. Also, most of the conventional objective metrics such as echo return loss enhancement (ERLE) and perceptual evaluation of speech quality (PESQ) do not correlate well with subjective speech quality tests in the presence of background noise and reverberation found in realistic environments. In this challenge, we open source two large datasets to train AEC models under both single talk and double talk scenarios. These datasets consist of recordings from more than 2,500 real audio devices and human speakers in real environments, as well as a synthetic dataset. We open source two large test sets, and we open source an online subjective test framework for researchers to quickly test their results. The winners of this challenge will be selected based on the average Mean Opinion Score (MOS) achieved across all different single talk and double talk scenarios.

Tanel Pärnamaa, Ando Saabas

Personalized speech enhancement (PSE) models can improve the audio quality of teleconferencing systems by adapting to the characteristics of a speaker's voice. However, most existing methods require a separate speaker embedding model to extract a vector representation of the speaker from enrollment audio, which adds complexity to the training and deployment process. We propose to use the internal representation of the PSE model itself as the speaker embedding, thereby avoiding the need for a separate model. We show that our approach performs equally well or better than the standard method of using a pre-trained speaker embedding model on noise suppression and echo cancellation tasks. Moreover, our approach surpasses the ICASSP 2023 Deep Noise Suppression Challenge winner by 0.15 in Mean Opinion Score.

Sungjin Ahn, Heeyoul Choi, Tanel Pärnamaa et al.

Current language models have a significant limitation in the ability to encode and decode factual knowledge. This is mainly because they acquire such knowledge from statistical co-occurrences although most of the knowledge words are rarely observed. In this paper, we propose a Neural Knowledge Language Model (NKLM) which combines symbolic knowledge provided by the knowledge graph with the RNN language model. By predicting whether the word to generate has an underlying fact or not, the model can generate such knowledge-related words by copying from the description of the predicted fact. In experiments, we show that the NKLM significantly improves the performance while generating a much smaller number of unknown words.