Learnable Frontends that do not Learn: Quantifying Sensitivity to Filterbank Initialisation

While much of modern speech and audio processing relies on deep neural networks trained using fixed audio representations, recent studies suggest great potential in acoustic frontends learnt jointly with a backend. In this study, we focus specifically on learnable filterbanks. Prior studies have reported that in frontends using learnable filterbanks initialised to a mel scale, the learned filters do not differ substantially from their initialisation. Using a Gabor-based filterbank, we investigate the sensitivity of a learnable filterbank to its initialisation using several initialisation strategies on two audio tasks: voice activity detection and bird species identification. We use the Jensen-Shannon Distance and analysis of the learned filters before and after training. We show that although performance is overall improved, the filterbanks exhibit strong sensitivity to their initialisation strategy. The limited movement from initialised values suggests that alternate optimisation strategies may allow a learnable frontend to reach better overall performance.

Learnable Acoustic Frontends in Bird Activity Detection

Autonomous recording units and passive acoustic monitoring present minimally intrusive methods of collecting bioacoustics data. Combining this data with species agnostic bird activity detection systems enables the monitoring of activity levels of bird populations. Unfortunately, variability in ambient noise levels and subject distance contribute to difficulties in accurately detecting bird activity in recordings. The choice of acoustic frontend directly affects the impact these issues have on system performance. In this paper, we benchmark traditional fixed-parameter acoustic frontends against the new generation of learnable frontends on a wide-ranging bird audio detection task using data from the DCASE2018 BAD Challenge. We observe that Per-Channel Energy Normalization is the best overall performer, achieving an accuracy of 89.9%, and that in general learnable frontends significantly outperform traditional methods. We also identify challenges in learning filterbanks for bird audio.

Low Resource Species Agnostic Bird Activity Detection

This paper explores low resource classifiers and features for the detection of bird activity, suitable for embedded Automatic Recording Units which are typically deployed for long term remote monitoring of bird populations. Features include low-level spectral parameters, statistical moments on pitch samples, and features derived from amplitude modulation. Performance is evaluated on several lightweight classifiers using the NIPS4Bplus dataset. Our experiments show that random forest classifiers perform best on this task, achieving an accuracy of 0.721 and an F1-Score of 0.604. We compare the results of our system against both a Convolutional Neural Network based detector, and standard MFCC features. Our experiments show that we can achieve equal or better performance in most metrics using features and models with a smaller computational cost and which are suitable for edge deployment.

Bioacoustic Event Detection with prototypical networks and data augmentation

This report presents deep learning and data augmentation techniques used by a system entered into the Few-Shot Bioacoustic Event Detection for the DCASE2021 Challenge. The remit was to develop a few-shot learning system for animal (mammal and bird) vocalisations. Participants were tasked with developing a method that can extract information from five exemplar vocalisations, or shots, of mammals or birds and detect and classify sounds in field recordings. In the system described in this report, prototypical networks are used to learn a metric space, from which classification is performed by computing the distance of a query point to class prototypes, classifying based on shortest distance. We describe the architecture of this network, feature extraction methods, and data augmentation performed on the given dataset and compare our work to the challenge's baseline networks.

AV Taris: Online Audio-Visual Speech Recognition

In recent years, Automatic Speech Recognition (ASR) technology has approached human-level performance on conversational speech under relatively clean listening conditions. In more demanding situations involving distant microphones, overlapped speech, background noise, or natural dialogue structures, the ASR error rate is at least an order of magnitude higher. The visual modality of speech carries the potential to partially overcome these challenges and contribute to the sub-tasks of speaker diarisation, voice activity detection, and the recovery of the place of articulation, and can compensate for up to 15dB of noise on average. This article develops AV Taris, a fully differentiable neural network model capable of decoding audio-visual speech in real time. We achieve this by connecting two recently proposed models for audio-visual speech integration and online speech recognition, namely AV Align and Taris. We evaluate AV Taris under the same conditions as AV Align and Taris on one of the largest publicly available audio-visual speech datasets, LRS2. Our results show that AV Taris is superior to the audio-only variant of Taris, demonstrating the utility of the visual modality to speech recognition within the real time decoding framework defined by Taris. Compared to an equivalent Transformer-based AV Align model that takes advantage of full sentences without meeting the real-time requirement, we report an absolute degradation of approximately 3% with AV Taris. As opposed to the more popular alternative for online speech recognition, namely the RNN Transducer, Taris offers a greatly simplified fully differentiable training pipeline. As a consequence, AV Taris has the potential to popularise the adoption of Audio-Visual Speech Recognition (AVSR) technology and overcome the inherent limitations of the audio modality in less optimal listening conditions.

Deep Multi-Scale Feature Learning for Defocus Blur Estimation

This paper presents an edge-based defocus blur estimation method from a single defocused image. We first distinguish edges that lie at depth discontinuities (called depth edges, for which the blur estimate is ambiguous) from edges that lie at approximately constant depth regions (called pattern edges, for which the blur estimate is well-defined). Then, we estimate the defocus blur amount at pattern edges only, and explore an interpolation scheme based on guided filters that prevents data propagation across the detected depth edges to obtain a dense blur map with well-defined object boundaries. Both tasks (edge classification and blur estimation) are performed by deep convolutional neural networks (CNNs) that share weights to learn meaningful local features from multi-scale patches centered at edge locations. Experiments on naturally defocused images show that the proposed method presents qualitative and quantitative results that outperform state-of-the-art (SOTA) methods, with a good compromise between running time and accuracy.

Learning to Count Words in Fluent Speech enables Online Speech Recognition

Sequence to Sequence models, in particular the Transformer, achieve state of the art results in Automatic Speech Recognition. Practical usage is however limited to cases where full utterance latency is acceptable. In this work we introduce Taris, a Transformer-based online speech recognition system aided by an auxiliary task of incremental word counting. We use the cumulative word sum to dynamically segment speech and enable its eager decoding into words. Experiments performed on the LRS2 and LibriSpeech datasets, of unconstrained and read speech respectively, show that the online system performs on a par with the offline one, while having a dynamic algorithmic delay of 5 segments. Furthermore, we show that the estimated segment length distribution resembles the word length distribution obtained with forced alignment, although our system does not require an exact segment-to-word equivalence. Taris introduces a negligible overhead compared to a standard Transformer, while the local relationship modelling between inputs and outputs grants invariance to sequence length by design.

Should we hard-code the recurrence concept or learn it instead ? Exploring the Transformer architecture for Audio-Visual Speech Recognition

The audio-visual speech fusion strategy AV Align has shown significant performance improvements in audio-visual speech recognition (AVSR) on the challenging LRS2 dataset. Performance improvements range between 7% and 30% depending on the noise level when leveraging the visual modality of speech in addition to the auditory one. This work presents a variant of AV Align where the recurrent Long Short-term Memory (LSTM) computation block is replaced by the more recently proposed Transformer block. We compare the two methods, discussing in greater detail their strengths and weaknesses. We find that Transformers also learn cross-modal monotonic alignments, but suffer from the same visual convergence problems as the LSTM model, calling for a deeper investigation into the dominant modality problem in machine learning.

Neural Generation of Dialogue Response Timings

The timings of spoken response offsets in human dialogue have been shown to vary based on contextual elements of the dialogue. We propose neural models that simulate the distributions of these response offsets, taking into account the response turn as well as the preceding turn. The models are designed to be integrated into the pipeline of an incremental spoken dialogue system (SDS). We evaluate our models using offline experiments as well as human listening tests. We show that human listeners consider certain response timings to be more natural based on the dialogue context. The introduction of these models into SDS pipelines could increase the perceived naturalness of interactions.

How to Teach DNNs to Pay Attention to the Visual Modality in Speech Recognition

Audio-Visual Speech Recognition (AVSR) seeks to model, and thereby exploit, the dynamic relationship between a human voice and the corresponding mouth movements. A recently proposed multimodal fusion strategy, AV Align, based on state-of-the-art sequence to sequence neural networks, attempts to model this relationship by explicitly aligning the acoustic and visual representations of speech. This study investigates the inner workings of AV Align and visualises the audio-visual alignment patterns. Our experiments are performed on two of the largest publicly available AVSR datasets, TCD-TIMIT and LRS2. We find that AV Align learns to align acoustic and visual representations of speech at the frame level on TCD-TIMIT in a generally monotonic pattern. We also determine the cause of initially seeing no improvement over audio-only speech recognition on the more challenging LRS2. We propose a regularisation method which involves predicting lip-related Action Units from visual representations. Our regularisation method leads to better exploitation of the visual modality, with performance improvements between 7% and 30% depending on the noise level. Furthermore, we show that the alternative Watch, Listen, Attend, and Spell network is affected by the same problem as AV Align, and that our proposed approach can effectively help it learn visual representations. Our findings validate the suitability of the regularisation method to AVSR and encourage researchers to rethink the multimodal convergence problem when having one dominant modality.