Polyphonic sound event localization and detection (SELD), which jointly performs sound event detection (SED) and direction-of-arrival (DoA) estimation, detects the type and occurrence time of sound events as well as their corresponding DoA angles simultaneously. We study the SELD task from a multi-task learning perspective. Two open problems are addressed in this paper. Firstly, to detect overlapping sound events of the same type but with different DoAs, we propose to use a trackwise output format and solve the accompanying track permutation problem with permutation-invariant training. Multi-head self-attention is further used to separate tracks. Secondly, a previous finding is that, by using hard parameter-sharing, SELD suffers from a performance loss compared with learning the subtasks separately. This is solved by a soft parameter-sharing scheme. We term the proposed method as Event Independent Network V2 (EINV2), which is an improved version of our previously-proposed method and an end-to-end network for SELD. We show that our proposed EINV2 for joint SED and DoA estimation outperforms previous methods by a large margin, and has comparable performance to state-of-the-art ensemble models. Index Terms— Sound event localization and detection, direction of arrival, event-independent, permutation-invariant training, multi-task learning.
Environmental audio tagging is a newly proposed task to predict the presence or absence of a specific audio event in a chunk. Deep neural network (DNN) based methods have been successfully adopted for predicting the audio tags in the domestic audio scene. In this paper, we propose to use a convolutional neural network (CNN) to extract robust features from mel-filter banks (MFBs), spectrograms or even raw waveforms for audio tagging. Gated recurrent unit (GRU) based recurrent neural networks (RNNs) are then cascaded to model the long-term temporal structure of the audio signal. To complement the input information, an auxiliary CNN is designed to learn on the spatial features of stereo recordings. We evaluate our proposed methods on Task 4 (audio tagging) of the Detection and Classification of Acoustic Scenes and Events 2016 (DCASE 2016) challenge. Compared with our recent DNN-based method, the proposed structure can reduce the equal error rate (EER) from 0.13 to 0.11 on the development set. The spatial features can further reduce the EER to 0.10. The performance of the end-to-end learning on raw waveforms is also comparable. Finally, on the evaluation set, we get the state-of-the-art performance with 0.12 EER while the performance of the best existing system is 0.15 EER.
The DCASE Challenge 2016 contains tasks for Acoustic Scene Classification (ASC), Acoustic Event Detection (AED), and audio tagging. Since 2006, Deep Neural Networks (DNNs) have been widely applied to computer visions, speech recognition and natural language processing tasks. In this paper, we provide DNN baselines for the DCASE Challenge 2016. In Task 1 we obtained accuracy of 81.0% using Mel + DNN against 77.2% by using Mel Frequency Cepstral Coefficients (MFCCs) + Gaussian Mixture Model (GMM). In Task 2 we obtained F value of 12.6% using Mel + DNN against 37.0% by using Constant Q Transform (CQT) + Nonnegative Matrix Factorization (NMF). In Task 3 we obtained F value of 36.3% using Mel + DNN against 23.7% by using MFCCs + GMM. In Task 4 we obtained Equal Error Rate (ERR) of 18.9% using Mel + DNN against 20.9% by using MFCCs + GMM. Therefore the DNN improves the baseline in Task 1, 3, and 4, although it is worse than the baseline in Task 2. This indicates that DNNs can be successful in many of these tasks, but may not always perform better than the baselines.
Audio tagging aims to perform multi-label classification on audio chunks and it is a newly proposed task in the Detection and Classification of Acoustic Scenes and Events 2016 (DCASE 2016) challenge. This task encourages research efforts to better analyze and understand the content of the huge amounts of audio data on the web. The difficulty in audio tagging is that it only has a chunk-level label without a frame-level label. This paper presents a weakly supervised method to not only predict the tags but also indicate the temporal locations of the occurred acoustic events. The attention scheme is found to be effective in identifying the important frames while ignoring the unrelated frames. The proposed framework is a deep convolutional recurrent model with two auxiliary modules: an attention module and a localization module. The proposed algorithm was evaluated on the Task 4 of DCASE 2016 challenge. State-of-the-art performance was achieved on the evaluation set with equal error rate (EER) reduced from 0.13 to 0.11, compared with the convolutional recurrent baseline system.
Sound event detection (SED) aims to detect when and recognize what sound events happen in an audio clip. Many supervised SED algorithms rely on strongly labelled data which contains the onset and offset annotations of sound events. However, many audio tagging datasets are weakly labelled, that is, only the presence of the sound events is known, without knowing their onset and offset annotations. In this paper, we propose a time-frequency (T-F) segmentation framework trained on weakly labelled data to tackle the sound event detection and separation problem. In training, a segmentation mapping is applied on a T-F representation, such as log mel spectrogram of an audio clip to obtain T-F segmentation masks of sound events. The T-F segmentation masks can be used for separating the sound events from the background scenes in the time-frequency domain. Then a classification mapping is applied on the T-F segmentation masks to estimate the presence probabilities of the sound events. We model the segmentation mapping using a convolutional neural network and the classification mapping using a global weighted rank pooling (GWRP). In SED, predicted onset and offset times can be obtained from the T-F segmentation masks. As a byproduct, separated waveforms of sound events can be obtained from the T-F segmentation masks. We remixed the DCASE 2018 Task 1 acoustic scene data with the DCASE 2018 Task 2 sound events data. When mixing under 0 dB, the proposed method achieved F1 scores of 0.534, 0.398 and 0.167 in audio tagging, frame-wise SED and event-wise SED, outperforming the fully connected deep neural network baseline of 0.331, 0.237 and 0.120, respectively. In T-F segmentation, we achieved an F1 score of 0.218, where previous methods were not able to do T-F segmentation.
Audio tagging is the task of predicting the presence or absence of sound classes within an audio clip. Previous work in audio tagging focused on relatively small datasets limited to recognising a small number of sound classes. We investigate audio tagging on AudioSet, which is a dataset consisting of over 2 million audio clips and 527 classes. AudioSet is weakly labelled, in that only the presence or absence of sound classes is known for each clip, while the onset and offset times are unknown. To address the weakly-labelled audio tagging problem, we propose attention neural networks as a way to attend the most salient parts of an audio clip. We bridge the connection between attention neural networks and multiple instance learning (MIL) methods, and propose decision-level and feature-level attention neural networks for audio tagging. We investigate attention neural networks modelled by different functions, depths and widths. Experiments on AudioSet show that the feature-level attention neural network achieves a state-of-the-art mean average precision (mAP) of 0.369, outperforming the best multiple instance learning (MIL) method of 0.317 and Google’s deep neural network baseline of 0.314. In addition, we discover that the audio tagging performance on AudioSet embedding features has a weak correlation with the number of training examples and the quality of labels of each sound class.
In this paper, we propose a multi-level attention model for the weakly labelled audio classification problem. The objective of audio classification is to predict the presence or the absence of sound events in an audio clip. Recently, Google published a large scale weakly labelled AudioSet dataset containing 2 million audio clips with only the presence or the absence labels of the sound events, without the onset and offset time of the sound events. Previously proposed attention models only applied a single attention module on the last layer of a neural network which limited the capacity of the attention model. In this paper, we propose a multi-level attention model which consists of multiple attention modules applied on the intermediate neural network layers. The outputs of these attention modules are concatenated to a vector followed by a fully connected layer to obtain the final prediction of each class. Experiments show that the proposed multi-attention attention model achieves a state-of-the-art mean average precision (mAP) of 0.360, outperforming the single attention model and the Google baseline system of 0.327 and 0.314, respectively.
Acoustic monitoring of bird species is an increasingly important field in signal processing. Many available bird sound datasets do not contain exact timestamp of the bird call but have a coarse weak label instead. Traditional Non-negative Matrix Factorization (NMF) models are not well designed to deal with weakly labeled data. In this paper we propose a novel Masked Non-negative Matrix Factorization (Masked NMF) approach for bird detection using weakly labeled data. During dictionary extraction we introduce a binary mask on the activation matrix. In that way we are able to control which parts of dictionary are used to reconstruct the training data. We compare our method with conventional NMF approaches and current state of the art methods. The proposed method outperforms the NMF baseline and offers a parsimonious model for bird detection on weakly labeled data. Moreover, to our knowledge, the proposed Masked NMF achieved the best result among non-deep learning methods on a test dataset used for the recent Bird Audio Detection Challenge.
Source separation is the task of separating an audio recording into individual sound sources. Source separation is fundamental for computational auditory scene analysis. Previous work on source separation has focused on separating particular sound classes such as speech and music. Much previous work requires mixtures and clean source pairs for training. In this work, we propose a source separation framework trained with weakly labelled data. Weakly labelled data only contains the tags of an audio clip, without the occurrence time of sound events. We first train a sound event detection system with AudioSet. The trained sound event detection system is used to detect segments that are most likely to contain a target sound event. Then a regression is learnt from a mixture of two randomly selected segments to a target segment conditioned on the audio tagging prediction of the target segment. Our proposed system can separate 527 kinds of sound classes from AudioSet within a single system. A U-Net is adopted for the separation system and achieves an average SDR of 5.67 dB over 527 sound classes in AudioSet.
The goal of Acoustic Scene Classification (ASC) is to recognise the environment in which an audio waveform has been recorded. Recently, deep neural networks have been applied to ASC and have achieved state-of-the-art performance. However, few works have investigated how to visualise and understand what a neural network has learnt from acoustic scenes. Previous work applied local pooling after each convolutional layer, therefore reduced the size of the feature maps. In this paper, we suggest that local pooling is not necessary, but the size of the receptive field is important. We apply atrous Convolutional Neural Networks (CNNs) with global attention pooling as the classification model. The internal feature maps of the attention model can be visualised and explained. On the Detection and Classification of Acoustic Scenes and Events (DCASE) 2018 dataset, our proposed method achieves an accuracy of 72.7 %, significantly outperforming the CNNs without dilation at 60.4 %. Furthermore, our results demonstrate that the learnt feature maps contain rich information on acoustic scenes in the time-frequency domain.
In this paper, we present a gated convolutional neural network and a temporal attention-based localization method for audio classification, which won the 1st place in the large-scale weakly supervised sound event detection task of Detection and Classification of Acoustic Scenes and Events (DCASE) 2017 challenge. The audio clips in this task, which are extracted from YouTube videos, are manually labelled with one or more audio tags, but without time stamps of the audio events, hence referred to as weakly labelled data. Two subtasks are defined in this challenge including audio tagging and sound event detection using this weakly labelled data. We propose a convolutional recurrent neural network (CRNN) with learnable gated linear units (GLUs) non-linearity applied on the log Mel spectrogram. In addition, we propose a temporal attention method along the frames to predict the locations of each audio event in a chunk from the weakly labelled data. The performances of our systems were ranked the 1st and the 2nd as a team in these two sub-tasks of DCASE 2017 challenge with F value 55.6% and Equal error 0.73, respectively.
Bird audio detection (BAD) aims to detect whether there is a bird call in an audio recording or not. One difficulty of this task is that the bird sound datasets are weakly labelled, that is only the presence or absence of a bird in a recording is known, without knowing when the birds call. We propose to apply joint detection and classification (JDC) model on the weakly labelled data (WLD) to detect and classify an audio clip at the same time. First, we apply VGG like convolutional neural network (CNN) on mel spectrogram as baseline. Then we propose a JDC-CNN model with VGG as a classifier and CNN as a detector. We report the denoising method including optimally-modified log-spectral amplitude (OM-LSA), median filter and spectral spectrogram will worse the classification accuracy on the contrary to previous work. JDC-CNN can predict the time stamps of the events from weakly labelled data, so is able to do sound event detection from WLD. We obtained area under curve (AUC) of 95.70% on the development data and 81.36% on the unseen evaluation data, which is nearly comparable to the baseline CNN model.
In supervised machine learning, the assumption that training data is labelled correctly is not always satisfied. In this paper, we investigate an instance of labelling error for classification tasks in which the dataset is corrupted with out-of-distribution (OOD) instances: data that does not belong to any of the target classes, but is labelled as such. We show that detecting and relabelling certain OOD instances, rather than discarding them, can have a positive effect on learning. The proposed method uses an auxiliary classifier, trained on data that is known to be in-distribution, for detection and relabelling. The amount of data required for this is shown to be small. Experiments are carried out on the FSDnoisy18k audio dataset, where OOD instances are very prevalent. The proposed method is shown to improve the performance of convolutional neural networks by a significant margin. Comparisons with other noise-robust techniques are similarly encouraging.
General-purpose audio tagging refers to classifying sounds that are of a diverse nature, and is relevant in many applications where domain-specific information cannot be exploited. The DCASE 2018 challenge introduces Task 2 for this very problem. In this task, there are a large number of classes and the audio clips vary in duration. Moreover, a subset of the labels are noisy. In this paper, we propose a system to address these challenges. The basis of our system is an ensemble of convolutional neural networks trained on log-scaled mel spectrograms. We use preprocessing and data augmentation methods to improve the performance further. To reduce the effects of label noise, two techniques are proposed: loss function weighting and pseudo-labeling. Experiments on the private test set of this task show that our system achieves state-of-the-art performance with a mean average precision score of 0.951
Audio tagging aims to assign one or several tags to an audio clip. Most of the datasets are weakly labelled, which means only the tags of the clip are known, without knowing the occurrence time of the tags. The labeling of an audio clip is often based on the audio events in the clip and no event level label is provided to the user. Previous works have used the bag of frames model assume the tags occur all the time, which is not the case in practice. We propose a joint detection-classification (JDC) model to detect and classify the audio clip simultaneously. The JDC model has the ability to attend to informative and ignore uninformative sounds. Then only informative regions are used for classification. Experimental results on the “CHiME Home” dataset show that the JDC model reduces the equal error rate (EER) from 19.0% to 16.9%. More interestingly, the audio event detector is trained successfully without needing the event level label.
Given binaural features as input, such as interaural level difference and interaural phase difference, Deep Neural Networks (DNNs) have been recently used to localize sound sources in a mixture of speech signals and/or noise, and to create time-frequency masks for the estimation of the sound sources in reverberant rooms. Here, we explore a more advanced system, where feed-forward DNNs are replaced by Convolutional Neural Networks (CNNs). In addition, the adjacent frames of each time frame (occurring before and after this frame) are used to exploit contextual information, thus improving the localization and separation for each source. The quality of the separation results is evaluated in terms of Signal to Distortion Ratio (SDR).
We propose a convolutional neural network (CNN) model based on an attention pooling method to classify ten different acoustic scenes, participating in the acoustic scene classiﬁcation task of the IEEE AASPChallengeonDetectionandClassiﬁcationofAcousticScenes and Events (DCASE 2018), which includes data from one device (subtask A) and data from three different devices (subtask B). The log mel spectrogram images of the audio waves are ﬁrst forwarded to convolutional layers, and then fed into an attention pooling layer to reduce the feature dimension and achieve classiﬁcation. From attention perspective, we build a weighted evaluation of the features, instead of simple max pooling or average pooling. On the ofﬁcial development set of the challenge, the best accuracy of subtask A is 72.6%,whichisanimprovementof12.9%whencomparedwiththe ofﬁcial baseline (p < .001 in a one-tailed z-test). For subtask B, the best result of our attention-based CNN is a signiﬁcant improvement of the baseline as well, in which the accuracies are 71.8%, 58.3%, and 58.3% for the three devices A to C (p < .001 for device A, p < .01 for device B, and p < .05 for device C).
Single-channel signal separation and deconvolution aims to separate and deconvolve individual sources from a single-channel mixture and is a challenging problem in which no prior knowledge of the mixing filters is available. Both individual sources and mixing filters need to be estimated. In addition, a mixture may contain non-stationary noise which is unseen in the training set. We propose a synthesizing-decomposition (S-D) approach to solve the single-channel separation and deconvolution problem. In synthesizing, a generative model for sources is built using a generative adversarial network (GAN). In decomposition, both mixing filters and sources are optimized to minimize the reconstruction error of the mixture. The proposed S-D approach achieves a peak-to-noise-ratio (PSNR) of 18.9 dB and 15.4 dB in image inpainting and completion, outperforming a baseline convolutional neural network PSNR of 15.3 dB and 12.2 dB, respectively and achieves a PSNR of 13.2 dB in source separation together with deconvolution, outperforming a convolutive non-negative matrix factorization (NMF) baseline of 10.1 dB.
Sound event detection (SED) and localization refer to recognizing sound events and estimating their spatial and temporal locations. Using neural networks has become the prevailing method for SED. In the area of sound localization, which is usually performed by estimating the direction of arrival (DOA), learning-based methods have recently been developed. In this paper, it is experimentally shown that the trained SED model is able to contribute to the direction of arrival estimation (DOAE). However, joint training of SED and DOAE degrades the performance of both. Based on these results, a two-stage polyphonic sound event detection and localization method is proposed. The method learns SED first, after which the learned feature layers are transferred for DOAE. It then uses the SED ground truth as a mask to train DOAE. The proposed method is evaluated on the DCASE 2019 Task 3 dataset, which contains different overlapping sound events in different environments. Experimental results show that the proposed method is able to improve the performance of both SED and DOAE, and also performs significantly better than the baseline method.
The detection of acoustic scenes is a challenging problem in which environmental sound events must be detected from a given audio signal. This includes classifying the events as well as estimating their onset and offset times. We approach this problem with a neural network architecture that uses the recentlyproposed capsule routing mechanism. A capsule is a group of activation units representing a set of properties for an entity of interest, and the purpose of routing is to identify part-whole relationships between capsules. That is, a capsule in one layer is assumed to belong to a capsule in the layer above in terms of the entity being represented. Using capsule routing, we wish to train a network that can learn global coherence implicitly, thereby improving generalization performance. Our proposed method is evaluated on Task 4 of the DCASE 2017 challenge. Results show that classification performance is state-of-the-art, achieving an Fscore of 58.6%. In addition, overfitting is reduced considerably compared to other architectures.
—Audio pattern recognition is an important research topic in the machine learning area, and includes several tasks such as audio tagging, acoustic scene classification, music classification , speech emotion classification and sound event detection. Recently, neural networks have been applied to tackle audio pattern recognition problems. However, previous systems are built on specific datasets with limited durations. Recently, in computer vision and natural language processing, systems pretrained on large-scale datasets have generalized well to several tasks. However, there is limited research on pretraining systems on large-scale datasets for audio pattern recognition. In this paper, we propose pretrained audio neural networks (PANNs) trained on the large-scale AudioSet dataset. These PANNs are transferred to other audio related tasks. We investigate the performance and computational complexity of PANNs modeled by a variety of convolutional neural networks. We propose an architecture called Wavegram-Logmel-CNN using both log-mel spectrogram and waveform as input feature. Our best PANN system achieves a state-of-the-art mean average precision (mAP) of 0.439 on AudioSet tagging, outperforming the best previous system of 0.392. We transfer PANNs to six audio pattern recognition tasks, and demonstrate state-of-the-art performance in several of those tasks. We have released the source code and pretrained models of PANNs: https://github.com/qiuqiangkong/audioset_tagging_cnn.
Acoustic Scene Classification (ASC) aims to classify the environment in which the audio signals are recorded. Recently, Convolutional Neural Networks (CNNs) have been successfully applied to ASC. However, the data distributions of the audio signals recorded with multiple devices are different. There has been little research on the training of robust neural networks on acoustic scene datasets recorded with multiple devices, and on explaining the operation of the internal layers of the neural networks. In this article, we focus on training and explaining device-robust CNNs on multi-device acoustic scene data. We propose conditional atrous CNNs with attention for multi-device ASC. Our proposed system contains an ASC branch and a device classification branch, both modelled by CNNs. We visualise and analyse the intermediate layers of the atrous CNNs. A time-frequency attention mechanism is employed to analyse the contribution of each time-frequency bin of the feature maps in the CNNs. On the Detection and Classification of Acoustic Scenes and Events (DCASE) 2018 ASC dataset, recorded with three devices, our proposed model performs significantly better than CNNs trained on single-device data.
Energy disaggregation, a.k.a. Non-Intrusive Load Monitoring, aims to separate the energy consumption of individual appliances from the readings of a mains power meter measuring the total energy consumption of, e.g. a whole house. Energy consumption of individual appliances can be useful in many applications, e.g., providing appliance-level feedback to the end users to help them understand their energy consumption and ultimately save energy. Recently, with the availability of large-scale energy consumption datasets, various neural network models such as convolutional neural networks and recurrent neural networks have been investigated to solve the energy disaggregation problem. Neural network models can learn complex patterns from large amounts of data and have been shown to outperform the traditional machine learning methods such as variants of hidden Markov models. However, current neural network methods for energy disaggregation are either computational expensive or are not capable of handling long-term dependencies. In this paper, we investigate the application of the recently developed WaveNet models for the task of energy disaggregation. Based on a real-world energy dataset collected from 20 households over two years, we show that WaveNet models outperforms the state-of-the-art deep learning methods proposed in the literature for energy disaggregation in terms of both error measures and computational cost. On the basis of energy disaggregation, we then investigate the performance of two deep-learning based frameworks for the task of on/off detection which aims at estimating whether an appliance is in operation or not. Based on the same dataset, we show that for the task of on/off detection the second framework, i.e., directly training a binary classifier, achieves better performance in terms of F1 score.
A method based on Deep Neural Networks (DNNs) and time-frequency masking has been recently developed for binaural audio source separation. In this method, the DNNs are used to predict the Direction Of Arrival (DOA) of the audio sources with respect to the listener which is then used to generate soft time-frequency masks for the recovery/estimation of the individual audio sources. In this paper, an algorithm called ‘dropout’ will be applied to the hidden layers, affecting the sparsity of hidden units activations: randomly selected neurons and their connections are dropped during the training phase, preventing feature co-adaptation. These methods are evaluated on binaural mixtures generated with Binaural Room Impulse Responses (BRIRs), accounting a certain level of room reverberation. The results show that the proposed DNNs system with randomly deleted neurons is able to achieve higher SDRs performances compared to the baseline method without the dropout algorithm.
Sound event detection (SED) methods typically rely on either strongly labelled data or weakly labelled data. As an alternative, sequentially labelled data (SLD) was proposed. In SLD, the events and the order of events in audio clips are known, without knowing the occurrence time of events. This paper proposes a connectionist temporal classification (CTC) based SED system that uses SLD instead of strongly labelled data, with a novel unsupervised clustering stage. Experiments on 41 classes of sound events show that the proposed two-stage method trained on SLD achieves performance comparable to the previous state-of-the-art SED system trained on strongly labelled data, and is far better than another state-of-the-art SED system trained on weakly labelled data, which indicates the effectiveness of the proposed two-stage method trained on SLD without any onset/offset time of sound events.
Acoustic scene generation (ASG) is a task to generate waveforms for acoustic scenes. ASG can be used to generate audio scenes for movies and computer games. Recently, neural networks such as SampleRNN have been used for speech and music generation. However, ASG is more challenging due to its wide variety. In addition, evaluating a generative model is also difficult. In this paper, we propose to use a conditional SampleRNN model to generate acoustic scenes conditioned on the input classes. We also propose objective criteria to evaluate the quality and diversity of the generated samples based on classification accuracy. The experiments on the DCASE 2016 Task 1 acoustic scene data show that with the generated audio samples, a classification accuracy of 65:5% can be achieved compared to samples generated by a random model of 6:7% and samples from real recording of 83:1%. The performance of a classifier trained only on generated samples achieves an accuracy of 51:3%, as opposed to an accuracy of 6:7% with samples generated by a random model.
In this technique report, we present a bunch of methods for the task 4 of Detection and Classification of Acoustic Scenes and Events 2017 (DCASE2017) challenge. This task evaluates systems for the large-scale detection of sound events using weakly labeled training data. The data are YouTube video excerpts focusing on transportation and warnings due to their industry applications. There are two tasks, audio tagging and sound event detection from weakly labeled data. Convolutional neural network (CNN) and gated recurrent unit (GRU) based recurrent neural network (RNN) are adopted as our basic framework. We proposed a learnable gating activation function for selecting informative local features. Attention-based scheme is used for localizing the specific events in a weakly-supervised mode. A new batch-level balancing strategy is also proposed to tackle the data unbalancing problem. Fusion of posteriors from different systems are found effective to improve the performance. In a summary, we get 61% F-value for the audio tagging subtask and 0.73 error rate (ER) for the sound event detection subtask on the development set. While the official multilayer perceptron (MLP) based baseline just obtained 13.1% F-value for the audio tagging and 1.02 for the sound event detection.
Sound event detection (SED) is a task to detect sound events in an audio recording. One challenge of the SED task is that many datasets such as the Detection and Classification of Acoustic Scenes and Events (DCASE) datasets are weakly labelled. That is, there are only audio tags for each audio clip without the onset and offset times of sound events. We compare segment-wise and clip-wise training for SED that is lacking in previous works. We propose a convolutional neural network transformer (CNN-Transfomer) for audio tagging and SED, and show that CNN-Transformer performs similarly to a convolutional recurrent neural network (CRNN). Another challenge of SED is that thresholds are required for detecting sound events. Previous works set thresholds empirically, and are not an optimal approaches. To solve this problem, we propose an automatic threshold optimization method. The first stage is to optimize the system with respect to metrics that do not depend on thresholds, such as mean average precision (mAP). The second stage is to optimize the thresholds with respect to metrics that depends on those thresholds. Our proposed automatic threshold optimization system achieves a state-of-the-art audio tagging F1 of 0.646, outperforming that without threshold optimization of 0.629, and a sound event detection F1 of 0.584, outperforming that without threshold optimization of 0.564.
Cardiovascular disease is one of the leading factors for death cause of human beings. In the past decade, heart sound classification has been increasingly studied for its feasibility to develop a non-invasive approach to monitor a subject’s health status. Particularly, relevant studies have benefited from the fast development of wearable devices and machine learning techniques. Nevertheless, finding and designing efficient acoustic properties from heart sounds is an expensive and time-consuming task. It is known that transfer learning methods can help extract higher representations automatically from the heart sounds without any human domain knowledge. However, most existing studies are based on models pre-trained on images, which may not fully represent the characteristics inherited from audio. To this end, we propose a novel transfer learning model pretrained on large scale audio data for a heart sound classification task. In this study, the PhysioNet CinC Challenge Dataset is used for evaluation. Experimental results demonstrate that, our proposed pre-trained audio models can outperform other popular models pre-trained by images by achieving the highest unweighted average recall at 89.7 %.
The Detection and Classiﬁcation of Acoustic Scenes and Events (DCASE) consists of ﬁve audio classiﬁcation and sound event detectiontasks: 1)Acousticsceneclassiﬁcation,2)General-purposeaudio tagging of Freesound, 3) Bird audio detection, 4) Weakly-labeled semi-supervised sound event detection and 5) Multi-channel audio classiﬁcation. In this paper, we create a cross-task baseline system for all ﬁve tasks based on a convlutional neural network (CNN): a “CNN Baseline” system. We implemented CNNs with 4 layers and 8 layers originating from AlexNet and VGG from computer vision. We investigated how the performance varies from task to task with the same conﬁguration of neural networks. Experiments show that deeper CNN with 8 layers performs better than CNN with 4 layers on all tasks except Task 1. Using CNN with 8 layers, we achieve an accuracy of 0.680 on Task 1, an accuracy of 0.895 and a mean average precision (MAP) of 0.928 on Task 2, an accuracy of 0.751 andanareaunderthecurve(AUC)of0.854onTask3,asoundevent detectionF1scoreof20.8%onTask4,andanF1scoreof87.75%on Task 5. We released the Python source code of the baseline systems under the MIT license for further research.
Audio tagging aims to detect the types of sound events occurring in an audio recording. To tag the polyphonic audio recordings, we propose to use Connectionist Temporal Classification (CTC) loss function on the top of Convolutional Recurrent Neural Network (CRNN) with learnable Gated Linear Units (GLUCTC), based on a new type of audio label data: Sequentially Labelled Data (SLD). In GLU-CTC, CTC objective function maps the frame-level probability of labels to clip-level probability of labels. To compare the mapping ability of GLU-CTC for sound events, we train a CRNN with GLU based on Global Max Pooling (GLU-GMP) and a CRNN with GLU based on Global Average Pooling (GLU-GAP). And we also compare the proposed GLU-CTC system with the baseline system, which is a CRNN trained using CTC loss function without GLU. The experiments show that the GLU-CTC achieves an Area Under Curve (AUC) score of 0.882 in audio tagging, outperforming the GLU-GMP of 0.803, GLU-GAP of 0.766 and baseline system of 0.837. That means based on the same CRNN model with GLU, the performance of CTC mapping is better than the GMP and GAP mapping. Given both based on the CTC mapping, the CRNN with GLU outperforms the CRNN without GLU.
This paper investigates the Audio Set classification. Audio Set is a large scale weakly labelled dataset (WLD) of audio clips. In WLD only the presence of a label is known, without knowing the happening time of the labels. We propose an attention model to solve this WLD problem and explain the attention model from a novel probabilistic perspective. Each audio clip in Audio Set consists of a collection of features. We call each feature as an instance and the collection as a bag following the terminology in multiple instance learning. In the attention model, each instance in the bag has a trainable probability measure for each class. The classification of the bag is the expectation of the classification output of the instances in the bag with respect to the learned probability measure. Experiments show that the proposed attention model achieves a mAP of 0.327 on Audio Set, outperforming the Google’s baseline of 0.314.
Audio tagging has attracted increasing attention since last decade and has various potential applications in many fields. The objective of audio tagging is to predict the labels of an audio clip. Recently deep learning methods have been applied to audio tagging and have achieved state-of-the-art performance, which provides a poor generalization ability on new data. However due to the limited size of audio tagging data such as DCASE data, the trained models tend to result in overfitting of the network. Previous data augmentation methods such as pitch shifting, time stretching and adding background noise do not show much improvement in audio tagging. In this paper, we explore the sample mixed data augmentation for the domestic audio tagging task, including mixup, SamplePairing and extrapolation. We apply a convolutional recurrent neural network (CRNN) with attention module with log-scaled mel spectrum as a baseline system. In our experiments, we achieve an state-of-the-art of equal error rate (EER) of 0.10 on DCASE 2016 task4 dataset with mixup approach, outperforming the baseline system without data augmentation.
Source separation (SS) aims to separate individual sources from an audio recording. Sound event detection (SED) aims to detect sound events from an audio recording. We propose a joint separation-classification (JSC) model trained only on weakly labelled audio data, that is, only the tags of an audio recording are known but the time of the events are unknown. First, we propose a separation mapping from the time-frequency (T-F) representation of an audio to the T-F segmentation masks of the audio events. Second, a classification mapping is built from each T-F segmentation mask to the presence probability of each audio event. In the source separation stage, sources of audio events and time of sound events can be obtained from the T-F segmentation masks. The proposed method achieves an equal error rate (EER) of 0.14 in SED, outperforming deep neural network baseline of 0.29. Source separation SDR of 8.08 dB is obtained by using global weighted rank pooling (GWRP) as probability mapping, outperforming the global max pooling (GMP) based probability mapping giving SDR at 0.03 dB. Source code of our work is published.
While deep learning is undoubtedly the predominant learning technique across speech processing, it is still not widely used in health-based applications. The corpora available for health-style recognition problems are often small, both concerning the total amount of data available and the number of individuals present. The Bipolar Disorder corpus, used in the 2018 Audio/Visual Emotion Challenge, contains only 218 audio samples from 46 individuals. Herein, we present a multi-instance learning framework aimed at constructing more reliable deep learning-based models in such conditions. First, we segment the speech files into multiple chunks. However, the problem is that each of the individual chunks is weakly labelled, as they are annotated with the label of the corresponding speech file, but may not be indicative of that label. We then train the deep learning-based (ensemble) multi-instance learning model, aiming at solving such a weakly labelled problem. The presented results demonstrate that this approach can improve the accuracy of feedforward, recurrent, and convolutional neural nets on the 3-class mania classification tasks undertaken on the Bipolar Disorder corpus.