位置：立项数据库 > 立项详情页
行车环境听觉模型及声音处理关键技术
  • 项目名称：行车环境听觉模型及声音处理关键技术
  • 项目类别：重大研究计划
  • 批准号：91120303
  • 申请代码：F010302
  • 项目来源：国家自然科学基金
  • 研究期限：2012-01-01-2015-12-31
  • 项目负责人：韩纪庆
  • 负责人职称：教授
  • 依托单位：哈尔滨工业大学
  • 批准年度：2011
中文摘要：

课题将围绕本重大研究计划的总体目标，根据本年度项目指南中的研究方向，开展行驶中无人驾驶车辆对车内外声音的自动检测、实时识别及理解方面的关键技术研究，为无人车的智能行为决策提供辅助的听觉信息。课题拟重点解决如下关键问题环境变化复杂情况下声音信号的有效拾取、噪声消除及声源定位；基于听觉认知的鲁棒声音特征提取及有效声音感知与环境状况理解；交通广播语音识别和信息抽取；并在此基础上构建行车环境下的听觉模型。通过课题的研究，拟提出一系列行车环境下普适的自动声音感知与理解的理论与方法；并借助无人车这一典型应用背景，探索特定场景和特定目标下听觉信息的认知机理。力争使研究成果达到国际先进水平，提升我国在视听觉信息处理领域的整体研究实力。

中文主题词： 声学事件检测；声音场景识别；声音感知与理解；麦克风阵列；语音信息抽取
英文摘要：

acoustics event detection；sound scene recognition；sound perception and understanding；microphone array；speech information extraction

英文主题词： acoustics event detection；sound scene recognition；sound perception and understanding；microphone array；speech information extraction
结论摘要：

本项目围绕重大研究计划的总体目标，按照研究计划书设定的研究内容，开展了无人车行车环境下车内外声音的自动检测、实时识别及理解方面的关键技术研究。为无人车的智能行为决策提供辅助的听觉信息。 在项目的实施过程中，不仅解决了计划书中所提出的关键问题，而且还在具体研究过程中扩展了相关的研究。在基于麦克风阵列的声源定位和信号分离方面，首先设计并制作了多通道麦克风阵列硬件设备；其次提出了基于变换域分析的声源定位算法；接着提出了基于隐马尔科夫随机场关联建模的信号分离方法。在基于麦克风阵列的噪声消除和信号增强方面，提出了基于深度神经网络后滤波的麦克风阵列噪声消除框架，能在消除噪声的同时，有效控制信号畸变。在基于耳蜗非线性与主动增益特性的鲁棒特征提取方面，提出了一种能有效仿真耳蜗非线性处理机制的声学特征提取方法，其鲁棒性明显优于传统的声学特征。在车辆周边及车内声学事件检测方面，首先提出了基于基频的声音分割方法；其次对基于高斯混合模型的建模方法进行了改进，提出了伪高斯混合模型方法；接着提出了异质混合模型方法，以适应实际应用中多种不同数据分布的建模问题。最后，提出了一种基于多尺度RBF核SVM的声学事件检测方法。在基于声音事件序列的场景识别方面，首先提出了基于低秩矩阵的特征表示方法和基于低秩张量的特征表示方法；其次提出了基于低秩支持向量机的音频场景识别方法；接着，提出了基于背景声的最小噪声统计量来识别音频场景的方法。在交通广播提示语音的识别与信息抽取方面，设计并实现了一个服务于无人车的交通广播语音识别与抽取系统。在无人车声音感知和理解系统构建方面，提出了一般化的无人车听觉能力模型和实现该听觉能力的技术框架。同时，构建了行车环境下的声音感知和理解原型系统，并在实际无人车平台上进行了测试，各项功能指标和性能指标达到要求。 项目组共发表或录用学术论文66篇，其中18篇进入SCI检索源，64篇进入EI检索源，申请专利11项，与多个国家的学者开展了学术交流与合作。培养博士研究生17名，硕士研究生20名。由2013年清华出版社出书一部。另一本专门讨论声学事件检测理论与方法的专著即将在2016年由科学出版社出版。

成果综合统计
成果类型
数量
  • 期刊论文
  • 会议论文
  • 专利
  • 获奖
  • 著作
  • 41
  • 55
  • 0
  • 0
  • 2
期刊论文
采用听觉滤波器的宽带MUSIC声源定位方法
汉语和英语音高重音自动标注法方法的对比分析
Integrating Binary Mask Estimation with MRF Priors of Cochleagram for Speech Separation
Statistical voice activity detection based on sparse representation over learned dictionary
基于噪音追踪的二值时频掩蔽到实值掩蔽的泛化算法
Auditory filter based broadband MUSIC algorithm for sound source localization
Optimization of Learned Dictionary for Sparse Coding in Speech Processing
A signal subspace dimension estimator based on F-norm with application to subspace-based multi-chann
The optimal ratio time-frequency mask for speech separation in terms of the signal-to-noise ratio.
The analysis of the simplification from the ideal ratio to binary mask in signal-to-noise ratio sens
A new Bayesian method incorporating with local correlation for IBM estimation
Latent topic model for audio retrieval
A novel signal subspace dimension estimator based on F-norm with application to subspace-based multi
Sparse Representation with Optimized Learned Dictionary for Robust Voice Activity Detection
A new framework for robust speech recognition in complex channel environments
Noise Robust Direction of Arrival Estimation for Speech Source With Weighted Bispectrum Spatial Corr
Spectrum enhancement with sparse coding for robust speech recognition
Dictionary evaluation and optimization for sparse coding based speech processing
Confidence Measure Based on Context Consistency Using Word Occurrence Probability and Topic Adaptati
Soft Margin Based Low-Rank Audio Signal Classification
行车噪声环境下基于人耳频率选择特性的声学特征提取方法
Identification of Objectionable Audio Segments Based on Pseudo and Heterogeneous Mixture Models
Fast Audio Retrieval Using Symbolized LSH Address Based on p-stable Distribution
基于p-稳定分布局部敏感哈希地址的鲁棒音频检索方法
融合引导概率的语音识别解码算法研究
SPARSE BASED AUDITORY MODEL FOR ROBUST SPEAKER RECOGNITION
Integrating Induced Probability into Decoding for Large Vocabulary Continuous Speech Recognition
基于深度学习语音分离技术的研究现状与进展
鲁棒声学事件检测综述
基于稀疏编码的鲁棒说话人识别
基于Fisher判别字典学习的说话人识别
会议论文
Noise-Robust Speaker Recognition Based on Morphological Component Analysis
Joint Optimization of Recurrent Networks Exploiting Source Auto-Regression for Source Separation
A new noise-tracking algorithm for generalizing binary time-frequency (T-F) masking to ratio masking
Low-rank audio signal classification under soft margin and trace norm constraints
A pairwise algorithm for pitch estimation and speech separation using deep stacking network
Cross-domain cooperative deep stacking network for speech separation
IMPROVED TONE MODELING BY EXPLOITING ARTICULATORY FEATURES FOR MANDARIN SPEECH RECOGNITION
Latent topic model based on Gaussian-LDA for audio retrieval
Monaural voiced speech separation with multipitch tracking
IMPROVED TONE MODELING BY EXPLOITING ARTICULATORY FEATURES FOR MANDARIN SPEECH RECOGNITION
UTOMATIC PROSODIC EVENTS DETECTION USING A TWO-STAGE SVM/CRF SEQUENCE CLASSIFIER WITH ACOUSTIC FEATU
声音感知在地面无人平台中的应用展望
Improving Deep Neural Networks Using Softplus Units
Case based reasoning solution to the problem of sustained learning in keyword spotting
Deep neural networks for Mandarin tone recognition
Deep stacking networks with time series for speech separation
Robust voice activity detection using the combination of short-term and long-term spectral patterns
DOA estimation of speech source in noisy environments with weighted spatial bispectrum correlation m
Direction of Arrival Estimation Based on Subband Weighting for Noisy Conditions
Multiscale collaborative speech denoising based on deep stacking network
行车环境下基于二值语谱图的声学事件检测
Weighted Spatial Bispectrum Correlation Matrix for DOA Estimation in the Presence of Interferences
Improving native accent identification using deep neural networks
Guarantees of Augmented Trace Norm Models in Tensor Recovery
Upper and Lower Bounds for Approximation of the Kullback-Leibler Divergence Between Hidden Markov Mo
Case Based Reasoning Solution to the Problem of Sustained Learning in Keyword Spotting
IMPROVED TONE MODELING BY EXPLOITING ARTICULATORY FEATURES FOR MANDARIN SPEECH RECOGNITION
IMPROVED TONE MODELING BY EXPLOITING ARTICULATORY FEATURES FOR MANDARIN SPEECH RECOGNITION
Direction of Arrival Estimation Based on Subband Weighting for Noisy Conditions
A Novel Confidence Measure Based on Context Consistency for Spoken Term Detection
An improved steady segment based decoding algorithm by using response probability for LVCSR
Speech fragment decoding techniques using silent pause detection
Sparse power spectrum based robust voice activity detector
Beam Pruning Based on Quantile for Keyword Spotting
A solution to residual noise in speech denoising with sparse representation
Automatic prosodic events detection using a two-stage SVM/CRF sequence classifier with acoustic feat
Multilingual Deep Neural Network for Music Genre Classification
Two-Stage Multi-Target Joint Learning for Monaural Speech Separation
Exploring Robustness of DNN/RNN for Extracting Speaker Baum-Welch Statistics in Mismatched Condition
Combing frame and segment based models for environmental sound classification
A novel confidence measure based on context consistency for spoken term detection
A Novel Codebook Representation Method and Encoding Strategy For Bag-of-Words Based Acoustic Event C
Attribute Knowledge Integration for Speech Recognition Based on Multi-Task Learning Neural Networks
Evaluation of dictionary for sparse coding in speech processing
Improving Deep Neural Networks Based Multi-Accent Mandarin Speech Recognition Using I-Vectors and Ac
Integration of articulatory knowledge and voicing features based on DNN/HMM for Mandarin speech reco
Hybrid SVM/HMM Architectures for Statistical Model-based Voice Activity Detection
DOA ESTIMATION OF SPEECH SOURCE IN NOISY ENVIRONMENTS WITH WEIGHTED SPATIAL BISPECTRUM CORRELATION M
DEEP STACKING NETWORKS WITH TIME SERIES FOR SPEECH SEPARATION
Evaluation of Dictionary for Sparse Coding in Speech Processing
Robust Minimum Statistics Project Coefficients Feature For Acoustic Environment Recognition
Learning Semantic Kernels For Scene Classification
IMPROVING DEEP NEURAL NETWORKS BY USING SPARSE DROPOUT STRATEGY
ENHANCED POWER-NORMALIZED FEATURES FOR MANDARIN ROBUST SPEECH RECOGNITION BASED ON A VOICED-UNVOICED
Interference Robust DOA Estimation of Human Speech by Exploiting Historical Information and Temporal
著作
语音信号处理（第二版）
语音信号处理（第二版）
相关项目
基于麦克风阵列的多说话人跟踪方法研究
期刊论文 47
基于麦克风阵列的多信道语音增强技术的研究
期刊论文 41 会议论文 1
两类FIR滤波器的最优设计
期刊论文 10 会议论文 2
鲁棒的非特定环境声学事件检测关键技术研究
期刊论文 14 会议论文 8 专利 3 著作 2
近场宽带麦克风阵列语音增强方法及阵列优化拓扑结构的研究
期刊论文 7 会议论文 3 专利 2 著作 1
稳健的麦克风阵列子带域波束形成算法研究
期刊论文 14 会议论文 5
基于麦克风阵列的移动机器人实时声源定位方法研究
期刊论文 5 会议论文 6
基于语音信号时频分布稀疏性的非平稳噪声抑制
期刊论文 6 会议论文 12
基于麦克风阵列的语音增强和定位方法研究
期刊论文 49 会议论文 11 著作 1
韩纪庆的项目
第七届信息科学联合会议
基于内在与潜在语义特征的声音段落级语义识别方法研究
期刊论文 2
数字音频水印系统的基本模型和算法研究
期刊论文 4
基于内容的音频信息检索关键技术研究
期刊论文 11 会议论文 6 专利 2 著作 1
基于子词网格的汉语语音检索关键技术研究
期刊论文 21 会议论文 3 著作 1
心理紧张和应力影响下Robust语音识别方法研究
期刊论文 2
鲁棒的非特定环境声学事件检测关键技术研究
期刊论文 14 会议论文 8 专利 3 著作 2