【1】 Analysis of Self-Attention Head Diversity for Conformer-based Automatic Speech Recognition
标题：基于仿真器的自动语音识别中的自注意头部差异性分析
链接：https://arxiv.org/abs/2209.06096
作者：Kartik Audhkhasi,Yinghui Huang,Bhuvana Ramabhadran,Pedro J. Moreno备注：Accepted for publication in Interspeech 2022摘要：注意层是现代端到端自动语音识别系统的组成部分，例如作为Transformer或Conformer体系结构的一部分。注意力通常是多头的，其中每个头具有独立的学习参数集，并且对相同的输入特征序列进行操作。多头注意的输出是各个头的输出的融合。我们实证分析了不同注意头产生的表征之间的差异，并证明了在训练过程中注意头变得高度相关。我们研究了几种提高注意头部多样性的方法，包括对每个头部使用不同的注意机制和辅助训练损失函数来提高头部多样性。实验表明，在训练过程中引入多样性促进辅助损失函数是一种更有效的方法，在Librispeech语料库上获得了高达6%的WER改善。最后，我们得出注意头的多样性与头部参数梯度的相似性之间的联系。摘要：Attention layers are an integral part of modern end-to-end automatic speech recognition systems, for instance as part of the Transformer or Conformer architecture. Attention is typically multi-headed, where each head has an independent set of learned parameters and operates on the same input feature sequence. The output of multi-headed attention is a fusion of the outputs from the individual heads. We empirically analyze the diversity between representations produced by the different attention heads and demonstrate that the heads become highly correlated during the course of training. We investigate a few approaches to increasing attention head diversity, including using different attention mechanisms for each head and auxiliary training loss functions to promote head diversity. We show that introducing diversity-promoting auxiliary loss functions during training is a more effective approach, and obtain WER improvements of up to 6% relative on the Librispeech corpus. Finally, we draw a connection between the diversity of attention heads and the similarity of the gradients of head parameters.【2】 Acoustic-Linguistic Features for Modeling Neurological Task Score in Alzheimer's
标题：阿尔茨海默病神经任务评分建模的声学-语言学特征
链接：https://arxiv.org/abs/2209.06085
作者：Saurav K. Aryal,Howard Prioleau,Legand Burge机构：EECS, Howard University, Washington, DC , USA备注：The paper has been accepted to Pacific Symposium on Biocomputing \c{opyright} [2022] World Scientific Publishing Co., Singapore, this http URL and is currently being camera-readied摘要：由于医疗技术、预防性保健的进步以及对老年健康的日益重视，全球平均预期寿命正在增加。因此，开发检测和跟踪老年人认知功能中衰老相关疾病的技术势在必行。特别地，考虑到阿尔茨海默病（AD）的流行和当前方法的成本，与该疾病的自动检测和评估相关的研究是关键的。由于AD影响语音和词汇的声学特性，自然语言处理和机器学习为可靠地检测AD提供了有前途的技术。我们比较和对比了10个线性回归模型在ADReSS挑战数据集上预测简易精神状态检查评分的性能。我们提取了13000多个手工制作和学习的特征，捕捉语言和声学现象。使用通过以下两种方法选择的54个顶级功能的子集：（1）递归消除和（2）相关性评分，我们在相同任务中的表现优于最先进的基线。在对每个模型的每个所选特征子集的统计显著性进行评分和评估后，我们发现，对于给定任务，手工制作的语言学特征比声学和习得的特征更显著。摘要：The average life expectancy is increasing globally due to advancements in medical technology, preventive health care, and a growing emphasis on gerontological health. Therefore, developing technologies that detect and track aging-associated disease in cognitive function among older adult populations is imperative. In particular, research related to automatic detection and evaluation of Alzheimer's disease (AD) is critical given the disease's prevalence and the cost of current methods. As AD impacts the acoustics of speech and vocabulary, natural language processing and machine learning provide promising techniques for reliably detecting AD. We compare and contrast the performance of ten linear regression models for predicting Mini-Mental Status Exam scores on the ADReSS challenge dataset. We extracted 13000+ handcrafted and learned features that capture linguistic and acoustic phenomena. Using a subset of 54 top features selected by two methods: (1) recursive elimination and (2) correlation scores, we outperform a state-of-the-art baseline for the same task. Upon scoring and evaluating the statistical significance of each of the selected subset of features for each model, we find that, for the given task, handcrafted linguistic features are more significant than acoustic and learned features.【3】 SongDriver: Real-time Music Accompaniment Generation without Logical Latency nor Exposure Bias
标题：SongDriver：没有逻辑延迟和曝光偏差的实时音乐伴奏生成
链接：https://arxiv.org/abs/2209.06054
作者：Zihao Wang,Kejun Zhang,Yuxing Wang,Chen Zhang,Qihao Liang,Pengfei Yu,Yongsheng Feng,Wenbo Liu,Yikai Wang,Yuntai Bao,Yiheng Yang机构：Hangzhou, China, Alibaba-Zhejiang University Joint, Institute of Frontier Technologies, Jingchu University of Technology, Hubei, China, Shandong University, Weihai, China备注：*Both Zihao Wang and Qihao Liang contribute equally to the paper and share the co-first authorship. This paper has been accepted by ACM Multimedia 2022 for oral presentation摘要：实时音乐伴奏生成在音乐教育、现场演出等音乐产业中有着广泛的应用。然而，自动实时音乐伴奏生成仍然未得到充分的研究，并且经常面临逻辑延迟和曝光偏差之间的权衡。本文提出了一种无逻辑延迟和曝光偏差的实时音乐伴奏生成系统SongDriver。具体地说，SongDriver将一个伴奏生成任务分成两个阶段：1）排列阶段，其中Transformer模型首先实时排列输入旋律的和弦，并缓存下一阶段的和弦，而不是将它们播放出来。2)在预测阶段，CRF模型根据先前缓存的和弦为即将到来的旋律生成可播放的多轨伴奏。通过这一两阶段策略，SongDriver直接为即将到来的旋律生成伴奏，实现零逻辑延迟。此外，当预测时间步长的和弦时，SongDriver会参考第一个相位的缓存和弦，而不是其先前的预测，这避免了曝光偏差问题。由于在实时条件下输入长度经常受到约束，另一个潜在的问题是长期顺序信息的丢失。为了弥补这一不足，我们从当前时间步之前的长期音乐片段中提取四个音乐特征作为全局信息。在实验中，我们在一些开源数据集和一个由中国风格的现代流行音乐乐谱构建的原始aiSong数据集上训练SongDriver。结果表明，SongDriver在客观和主观指标上均优于现有SOTA模型，同时显著降低了物理延迟。摘要：Real-time music accompaniment generation has a wide range of applications in the music industry, such as music education and live performances. However, automatic real-time music accompaniment generation is still understudied and often faces a trade-off between logical latency and exposure bias. In this paper, we propose SongDriver, a real-time music accompaniment generation system without logical latency nor exposure bias. Specifically, SongDriver divides one accompaniment generation task into two phases: 1) The arrangement phase, where a Transformer model first arranges chords for input melodies in real-time, and caches the chords for the next phase instead of playing them out. 2) The prediction phase, where a CRF model generates playable multi-track accompaniments for the coming melodies based on previously cached chords. With this two-phase strategy, SongDriver directly generates the accompaniment for the upcoming melody, achieving zero logical latency. Furthermore, when predicting chords for a timestep, SongDriver refers to the cached chords from the first phase rather than its previous predictions, which avoids the exposure bias problem. Since the input length is often constrained under real-time conditions, another potential problem is the loss of long-term sequential information. To make up for this disadvantage, we extract four musical features from a long-term music piece before the current time step as global information. In the experiment, we train SongDriver on some open-source datasets and an original \`aiSong Dataset built from Chinese-style modern pop music scores. The results show that SongDriver outperforms existing SOTA (state-of-the-art) models on both objective and subjective metrics, meanwhile significantly reducing the physical latency.【4】 A Distributed Acoustic Sensor System for Intelligent Transportation using Deep Learning
标题：基于深度学习的智能交通分布式声学传感系统
链接：https://arxiv.org/abs/2209.05978
作者：Chia-Yen Chiang,Mona Jaber,Peter Hayward机构：Jaber are with the School of Electronic Engineering andComputer Science, Queen Mary University of London摘要：智能交通系统（ITS）是可持续发展和绿色城市生活的关键。ITS是数据驱动的，由大量的传感器（从气动管道到智能相机）实现。本工作探索了一种基于光纤分布式声学传感器（DAS）的新型交通分析数据源。车辆类型的检测和车辆占用率的估计是智能交通系统（ITS）的主要研究内容。第一种是由跟踪、控制和预测交通流的需要所激发的。第二个目标是管制高占用率的车道，试图减少排放和拥堵。这些任务通常由检查车辆的个人或通过使用新兴的计算机视觉技术来执行。前者既不可扩展，效率也不高，而后者则会侵犯乘客的隐私。为此，我们提出了一种深度学习技术来分析DAS信号，以通过持续感知解决这一挑战，而不暴露个人信息。我们提出了一种深度学习方法来处理DAS信号，并基于在受控条件下收集的DAS数据实现了92%的车辆分类准确率和92-97%的占用检测。摘要：Intelligent transport systems (ITS) are pivotal in the development of sustainable and green urban living. ITS is data-driven and enabled by the profusion of sensors ranging from pneumatic tubes to smart cameras. This work explores a novel data source based on optical fibre-based distributed acoustic sensors (DAS) for traffic analysis. Detecting the type of vehicle and estimating the occupancy of vehicles are prime concerns in ITS. The first is motivated by the need for tracking, controlling, and forecasting traffic flow. The second targets the regulation of high occupancy vehicle lanes in an attempt to reduce emissions and congestion. These tasks are often conducted by individuals inspecting vehicles or through the use of emerging computer vision technologies. The former is not scale-able nor efficient whereas the latter is intrusive to passengers' privacy. To this end, we propose a deep learning technique to analyse DAS signals to address this challenge through continuous sensing and without exposing personal information. We propose a deep learning method for processing DAS signals and achieve 92% vehicle classification accuracy and 92-97% in occupancy detection based on DAS data collected under controlled conditions.【5】 Binaural Signal Representations for Joint Sound Event Detection and Acoustic Scene Classification
标题：用于联合声事件检测和声场景分类的双耳信号表示
链接：https://arxiv.org/abs/2209.05900
作者：Daniel Aleksander Krause,Annamaria Mesaros机构：Tampere University, Tampere, Finland摘要：声音事件检测（SED）和声学场景分类（ASC）是两个被广泛研究的音频任务，是声学场景分析研究的重要组成部分。考虑到声音事件和声学场景之间的共享信息，联合执行这两个任务是复杂机器听音系统的自然部分。在这篇文章中，我们研究了几个空间音频特征在训练执行SED和ASC的联合深度神经网络（DNN）模型中的有用性。在两个不同的双耳录音数据集上进行了实验，分析了单独或联合进行SED和ASC的差异。结果表明，与仅基于logmel能量的基线方法相比，使用特定的双耳特征，主要是广义互相关相位变换（GCC-phat）和相位差的正弦和余弦，在单独和联合任务中都产生了更好的性能模型。摘要：Sound event detection (SED) and Acoustic scene classification (ASC) are two widely researched audio tasks that constitute an important part of research on acoustic scene analysis. Considering shared information between sound events and acoustic scenes, performing both tasks jointly is a natural part of a complex machine listening system. In this paper, we investigate the usefulness of several spatial audio features in training a joint deep neural network (DNN) model performing SED and ASC. Experiments are performed for two different datasets containing binaural recordings and synchronous sound event and acoustic scene labels to analyse the differences between performing SED and ASC separately or jointly. The presented results show that the use of specific binaural features, mainly the Generalized Cross Correlation with Phase Transform (GCC-phat) and sines and cosines of phase differences, result in a better performing model in both separate and joint tasks as compared with baseline methods based on logmel energies only.【1】 Streaming End-to-End Multilingual Speech Recognition with Joint Language Identification
标题：采用联合语言识别的流端到端多语言语音识别
链接：https://arxiv.org/abs/2209.06058
作者：Chao Zhang,Bo Li,Tara Sainath,Trevor Strohman,Sepand Mavandadi,Shuo-yiin Chang,Parisa Haghani摘要：语言识别是自动语音识别（ASR）中许多下游任务的关键，将其作为一项附加任务集成到多语言端到端ASR中是非常有益的。本文提出了一种改进的基于级联编码器的递归神经网络传感器模型（RNN-T）的结构，该模型中集成了一个逐帧语言标识符（LID）预测器。具有级联编码器的RNN-T可以使用没有右上下文的第一遍解码实现低延迟的流式ASR，并且使用具有较长右上下文的第二遍解码实现较低的误字率（WER）。通过利用在正确上下文中的这种差异和统计池的流实现，所提出的方法可以以很少的额外测试时间成本来实现精确的流LID预测。在9个语言环境的语音搜索数据集上的实验结果表明，该方法的LID预测准确率平均为96.2%，二次遍历WER与在输入中包含oracle LID时的结果相同.摘要：Language identification is critical for many downstream tasks in automatic speech recognition (ASR), and is beneficial to integrate into multilingual end-to-end ASR as an additional task. In this paper, we propose to modify the structure of the cascaded-encoder-based recurrent neural network transducer (RNN-T) model by integrating a per-frame language identifier (LID) predictor. RNN-T with cascaded encoders can achieve streaming ASR with low latency using first-pass decoding with no right-context, and achieve lower word error rates (WERs) using second-pass decoding with longer right-context. By leveraging such differences in the right-contexts and a streaming implementation of statistics pooling, the proposed method can achieve accurate streaming LID prediction with little extra test-time cost. Experimental results on a voice search dataset with 9 language locales shows that the proposed method achieves an average of 96.2% LID prediction accuracy and the same second-pass WER as that obtained by including oracle LID in the input.【2】 Learning ASR pathways: A sparse multilingual ASR model
标题：学习ASR路径：一个稀疏的多语言ASR模型
链接：https://arxiv.org/abs/2209.05735
作者：Mu Yang,Andros Tjandra,Chunxi Liu,David Zhang,Duc Le,John H. L. Hansen,Ozlem Kalinli机构：University of Texas at Dallas, USA,Meta AI, USA摘要：神经网络剪枝可以有效地应用于自动语音识别（ASR）模型的压缩。然而，在多语言ASR中，执行与语言无关的修剪可能导致某些语言上的严重性能降级，因为与语言无关的修剪掩码可能不适合所有语言，并且丢弃重要的语言特定参数。在这项工作中，我们提出了ASR路径，一个稀疏的多语言ASR模型，激活语言特定的子网络（“路径”），使每种语言的参数被显式学习。通过重叠的子网络，共享参数还可以通过联合多语言培训实现低资源语言的知识转移。提出了一种新的ASR路径学习算法，并利用流式RNN-T模型在4种语言上对该算法进行了评估.我们提出的ASR路径优于密集模型（-5.0%平均WER）和语言不可知修剪模型（-21.4%平均WER），并且与单语稀疏模型相比，在低资源语言上提供了更好的性能。摘要：Neural network pruning can be effectively applied to compress automatic speech recognition (ASR) models. However, in multilingual ASR, performing language-agnostic pruning may lead to severe performance degradation on some languages because language-agnostic pruning masks may not fit all languages and discard important language-specific parameters. In this work, we present ASR pathways, a sparse multilingual ASR model that activates language-specific sub-networks ("pathways"), such that the parameters for each language are learned explicitly. With the overlapping sub-networks, the shared parameters can also enable knowledge transfer for lower resource languages via joint multilingual training. We propose a novel algorithm to learn ASR pathways, and evaluate the proposed method on 4 languages with a streaming RNN-T model. Our proposed ASR pathways outperform both dense models (-5.0% average WER) and a language-agnostically pruned model (-21.4% average WER), and provide better performance on low-resource languages compared to the monolingual sparse models.【3】 Analysis of Self-Attention Head Diversity for Conformer-based Automatic Speech Recognition
标题：基于仿真器的自动语音识别中的自注意头部差异性分析
链接：https://arxiv.org/abs/2209.06096
* 与cs.SD语音【1】为同一篇
作者：Kartik Audhkhasi,Yinghui Huang,Bhuvana Ramabhadran,Pedro J. Moreno备注：Accepted for publication in Interspeech 2022摘要：注意层是现代端到端自动语音识别系统的组成部分，例如作为Transformer或Conformer体系结构的一部分。注意力通常是多头的，其中每个头具有独立的学习参数集，并且对相同的输入特征序列进行操作。多头注意的输出是各个头的输出的融合。我们实证分析了不同注意头产生的表征之间的差异，并证明了在训练过程中注意头变得高度相关。我们研究了几种提高注意头部多样性的方法，包括对每个头部使用不同的注意机制和辅助训练损失函数来提高头部多样性。实验表明，在训练过程中引入多样性促进辅助损失函数是一种更有效的方法，在Librispeech语料库上获得了高达6%的WER改善。最后，我们得出注意头的多样性与头部参数梯度的相似性之间的联系。摘要：Attention layers are an integral part of modern end-to-end automatic speech recognition systems, for instance as part of the Transformer or Conformer architecture. Attention is typically multi-headed, where each head has an independent set of learned parameters and operates on the same input feature sequence. The output of multi-headed attention is a fusion of the outputs from the individual heads. We empirically analyze the diversity between representations produced by the different attention heads and demonstrate that the heads become highly correlated during the course of training. We investigate a few approaches to increasing attention head diversity, including using different attention mechanisms for each head and auxiliary training loss functions to promote head diversity. We show that introducing diversity-promoting auxiliary loss functions during training is a more effective approach, and obtain WER improvements of up to 6% relative on the Librispeech corpus. Finally, we draw a connection between the diversity of attention heads and the similarity of the gradients of head parameters.【4】 Acoustic-Linguistic Features for Modeling Neurological Task Score in Alzheimer's
标题：阿尔茨海默病神经任务评分建模的声学-语言学特征
链接：https://arxiv.org/abs/2209.06085
* 与cs.SD语音【2】为同一篇
作者：Saurav K. Aryal,Howard Prioleau,Legand Burge机构：EECS, Howard University, Washington, DC , USA备注：The paper has been accepted to Pacific Symposium on Biocomputing \c{opyright} [2022] World Scientific Publishing Co., Singapore, this http URL and is currently being camera-readied摘要：由于医疗技术、预防性保健的进步以及对老年健康的日益重视，全球平均预期寿命正在增加。因此，开发检测和跟踪老年人认知功能中衰老相关疾病的技术势在必行。特别地，考虑到阿尔茨海默病（AD）的流行和当前方法的成本，与该疾病的自动检测和评估相关的研究是关键的。由于AD影响语音和词汇的声学特性，自然语言处理和机器学习为可靠地检测AD提供了有前途的技术。我们比较和对比了10个线性回归模型在ADReSS挑战数据集上预测简易精神状态检查评分的性能。我们提取了13000多个手工制作和学习的特征，捕捉语言和声学现象。使用通过以下两种方法选择的54个顶级功能的子集：（1）递归消除和（2）相关性评分，我们在相同任务中的表现优于最先进的基线。在对每个模型的每个所选特征子集的统计显著性进行评分和评估后，我们发现，对于给定任务，手工制作的语言学特征比声学和习得的特征更显著。摘要：The average life expectancy is increasing globally due to advancements in medical technology, preventive health care, and a growing emphasis on gerontological health. Therefore, developing technologies that detect and track aging-associated disease in cognitive function among older adult populations is imperative. In particular, research related to automatic detection and evaluation of Alzheimer's disease (AD) is critical given the disease's prevalence and the cost of current methods. As AD impacts the acoustics of speech and vocabulary, natural language processing and machine learning provide promising techniques for reliably detecting AD. We compare and contrast the performance of ten linear regression models for predicting Mini-Mental Status Exam scores on the ADReSS challenge dataset. We extracted 13000+ handcrafted and learned features that capture linguistic and acoustic phenomena. Using a subset of 54 top features selected by two methods: (1) recursive elimination and (2) correlation scores, we outperform a state-of-the-art baseline for the same task. Upon scoring and evaluating the statistical significance of each of the selected subset of features for each model, we find that, for the given task, handcrafted linguistic features are more significant than acoustic and learned features.【5】 SongDriver: Real-time Music Accompaniment Generation without Logical Latency nor Exposure Bias
标题：SongDriver：没有逻辑延迟和曝光偏差的实时音乐伴奏生成
链接：https://arxiv.org/abs/2209.06054
* 与cs.SD语音【3】为同一篇
作者：Zihao Wang,Kejun Zhang,Yuxing Wang,Chen Zhang,Qihao Liang,Pengfei Yu,Yongsheng Feng,Wenbo Liu,Yikai Wang,Yuntai Bao,Yiheng Yang机构：Hangzhou, China, Alibaba-Zhejiang University Joint, Institute of Frontier Technologies, Jingchu University of Technology, Hubei, China, Shandong University, Weihai, China备注：*Both Zihao Wang and Qihao Liang contribute equally to the paper and share the co-first authorship. This paper has been accepted by ACM Multimedia 2022 for oral presentation摘要：实时音乐伴奏生成在音乐教育、现场演出等音乐产业中有着广泛的应用。然而，自动实时音乐伴奏生成仍然未得到充分的研究，并且经常面临逻辑延迟和曝光偏差之间的权衡。本文提出了一种无逻辑延迟和曝光偏差的实时音乐伴奏生成系统SongDriver。具体地说，SongDriver将一个伴奏生成任务分成两个阶段：1）排列阶段，其中Transformer模型首先实时排列输入旋律的和弦，并缓存下一阶段的和弦，而不是将它们播放出来。2)在预测阶段，CRF模型根据先前缓存的和弦为即将到来的旋律生成可播放的多轨伴奏。通过这一两阶段策略，SongDriver直接为即将到来的旋律生成伴奏，实现零逻辑延迟。此外，当预测时间步长的和弦时，SongDriver会参考第一个相位的缓存和弦，而不是其先前的预测，这避免了曝光偏差问题。由于在实时条件下输入长度经常受到约束，另一个潜在的问题是长期顺序信息的丢失。为了弥补这一不足，我们从当前时间步之前的长期音乐片段中提取四个音乐特征作为全局信息。在实验中，我们在一些开源数据集和一个由中国风格的现代流行音乐乐谱构建的原始aiSong数据集上训练SongDriver。结果表明，SongDriver在客观和主观指标上均优于现有SOTA模型，同时显著降低了物理延迟。摘要：Real-time music accompaniment generation has a wide range of applications in the music industry, such as music education and live performances. However, automatic real-time music accompaniment generation is still understudied and often faces a trade-off between logical latency and exposure bias. In this paper, we propose SongDriver, a real-time music accompaniment generation system without logical latency nor exposure bias. Specifically, SongDriver divides one accompaniment generation task into two phases: 1) The arrangement phase, where a Transformer model first arranges chords for input melodies in real-time, and caches the chords for the next phase instead of playing them out. 2) The prediction phase, where a CRF model generates playable multi-track accompaniments for the coming melodies based on previously cached chords. With this two-phase strategy, SongDriver directly generates the accompaniment for the upcoming melody, achieving zero logical latency. Furthermore, when predicting chords for a timestep, SongDriver refers to the cached chords from the first phase rather than its previous predictions, which avoids the exposure bias problem. Since the input length is often constrained under real-time conditions, another potential problem is the loss of long-term sequential information. To make up for this disadvantage, we extract four musical features from a long-term music piece before the current time step as global information. In the experiment, we train SongDriver on some open-source datasets and an original \`aiSong Dataset built from Chinese-style modern pop music scores. The results show that SongDriver outperforms existing SOTA (state-of-the-art) models on both objective and subjective metrics, meanwhile significantly reducing the physical latency.【6】 A Distributed Acoustic Sensor System for Intelligent Transportation using Deep Learning
标题：基于深度学习的智能交通分布式声学传感系统
链接：https://arxiv.org/abs/2209.05978
* 与cs.SD语音【4】为同一篇
作者：Chia-Yen Chiang,Mona Jaber,Peter Hayward机构：Jaber are with the School of Electronic Engineering andComputer Science, Queen Mary University of London摘要：智能交通系统（ITS）是可持续发展和绿色城市生活的关键。ITS是数据驱动的，由大量的传感器（从气动管道到智能相机）实现。本工作探索了一种基于光纤分布式声学传感器（DAS）的新型交通分析数据源。车辆类型的检测和车辆占用率的估计是智能交通系统（ITS）的主要研究内容。第一种是由跟踪、控制和预测交通流的需要所激发的。第二个目标是管制高占用率的车道，试图减少排放和拥堵。这些任务通常由检查车辆的个人或通过使用新兴的计算机视觉技术来执行。前者既不可扩展，效率也不高，而后者则会侵犯乘客的隐私。为此，我们提出了一种深度学习技术来分析DAS信号，以通过持续感知解决这一挑战，而不暴露个人信息。我们提出了一种深度学习方法来处理DAS信号，并基于在受控条件下收集的DAS数据实现了92%的车辆分类准确率和92-97%的占用检测。摘要：Intelligent transport systems (ITS) are pivotal in the development of sustainable and green urban living. ITS is data-driven and enabled by the profusion of sensors ranging from pneumatic tubes to smart cameras. This work explores a novel data source based on optical fibre-based distributed acoustic sensors (DAS) for traffic analysis. Detecting the type of vehicle and estimating the occupancy of vehicles are prime concerns in ITS. The first is motivated by the need for tracking, controlling, and forecasting traffic flow. The second targets the regulation of high occupancy vehicle lanes in an attempt to reduce emissions and congestion. These tasks are often conducted by individuals inspecting vehicles or through the use of emerging computer vision technologies. The former is not scale-able nor efficient whereas the latter is intrusive to passengers' privacy. To this end, we propose a deep learning technique to analyse DAS signals to address this challenge through continuous sensing and without exposing personal information. We propose a deep learning method for processing DAS signals and achieve 92% vehicle classification accuracy and 92-97% in occupancy detection based on DAS data collected under controlled conditions.【7】 Binaural Signal Representations for Joint Sound Event Detection and Acoustic Scene Classification
标题：用于联合声事件检测和声场景分类的双耳信号表示
链接：https://arxiv.org/abs/2209.05900
* 与cs.SD语音【5】为同一篇
作者：Daniel Aleksander Krause,Annamaria Mesaros机构：Tampere University, Tampere, Finland摘要：声音事件检测（SED）和声学场景分类（ASC）是两个被广泛研究的音频任务，是声学场景分析研究的重要组成部分。考虑到声音事件和声学场景之间的共享信息，联合执行这两个任务是复杂机器听音系统的自然部分。在这篇文章中，我们研究了几个空间音频特征在训练执行SED和ASC的联合深度神经网络（DNN）模型中的有用性。在两个不同的双耳录音数据集上进行了实验，分析了单独或联合进行SED和ASC的差异。结果表明，与仅基于logmel能量的基线方法相比，使用特定的双耳特征，主要是广义互相关相位变换（GCC-phat）和相位差的正弦和余弦，在单独和联合任务中都产生了更好的性能模型。摘要：Sound event detection (SED) and Acoustic scene classification (ASC) are two widely researched audio tasks that constitute an important part of research on acoustic scene analysis. Considering shared information between sound events and acoustic scenes, performing both tasks jointly is a natural part of a complex machine listening system. In this paper, we investigate the usefulness of several spatial audio features in training a joint deep neural network (DNN) model performing SED and ASC. Experiments are performed for two different datasets containing binaural recordings and synchronous sound event and acoustic scene labels to analyse the differences between performing SED and ASC separately or jointly. The presented results show that the use of specific binaural features, mainly the Generalized Cross Correlation with Phase Transform (GCC-phat) and sines and cosines of phase differences, result in a better performing model in both separate and joint tasks as compared with baseline methods based on logmel energies only.机器翻译，仅供参考
每日语音速递