多特征相结合的带噪语音端点检测算法的研究

提出了一种抗噪声的端点检测新方法。针对谱熵特征对清音的检测性能以及抗噪声性能较差的缺点,结合对清音检测性能较好的短时过零率特征,以及抗噪声性能良好的美尔倒谱距离特征,实现了基于多种特征相结合的抗噪声的语音端点检测。仿真实验表明,该方法能显著提高端点检测在高噪声环境下的检测性能。     

Vehicle recognition using acoustic sensor networks in complex scenes via kernel sparse representation

This paper proposes a method of vehicle recognition via kernel sparse representation using acoustic sensor networks in complex scenes. This algorithm uses the Mel frequency cepstral coefficients to extract the acoustic features of vehicles and maps them into a high-dimensional feature space with a kernel function to get linearly separable samples. After extending sparse representation to the kernel space and constructing the over-complete dictionary, the objective vehicles will be recognized by solving the optimization problem. Experiments show that the proposed algorithm gives good performance on vehicle recognition under the circumstance of complex data sets. Compared with other traditional acoustic classification algorithms, the method improves the precision of recognition.     

Artificial neural networks as speech recognisers for dysarthric speech: Identifying the best-performing set of MFCC parameters and studying a speaker-independent approach

Dysarthria is a neurological impairment of controlling the motor speech articulators that compromises the speech signal. Automatic Speech Recognition (ASR) can be very helpful for speakers with dysarthria because the disabled persons are often physically incapacitated. Mel-Frequency Cepstral Coefficients (MFCCs) have been proven to be an appropriate representation of dysarthric speech, but the question of which MFCC-based feature set represents dysarthric acoustic features most effectively has not been answered. Moreover, most of the current dysarthric speech recognisers are either speaker-dependent (SD) or speaker-adaptive (SA), and they perform poorly in terms of generalisability as a speaker-independent (SI) model. First, by comparing the results of 28 dysarthric SD speech recognisers, this study identifies the best-performing set of MFCC parameters, which can represent dysarthric acoustic features to be used in Artificial Neural Network (ANN)-based ASR. Next, this paper studies the application of ANNs as a fixed-length isolated-word SI ASR for individuals who suffer from dysarthria. The results show that the speech recognisers trained by the conventional 12 coefficients MFCC features without the use of delta and acceleration features provided the best accuracy, and the proposed SI ASR recognised the speech of the unforeseen dysarthric evaluation subjects with word recognition rate of 68.38%.     

噪音情境下生态环境声音的分类

提出一种对含有噪音的生态环境声音进行分类的方法.首先,匹配追踪(Matching Pursuit,简称MP)算法可以提取有效信号的时-频特征,减弱噪音的干扰.支持向量机(support Vector Machine,简称SVM)分类器的鲁棒性比较好,所以提出使用SVM基于MP时-频特征建立模型(简称MP-SVM)对含有噪音的生态环境声音进行分类.实验得出MP-SVM可取得较好的分类效果,证明了MP时-频特征和SVM分类器具有较好的抗噪性.     

视听觉跨模态表面材质检索

针对文本图像特征有时无法满足对物体材质进行真实准确分析的情况,本文在视听领域使用跨模态检索方法进行表面材质检索。首先提取声音的梅尔频率倒谱系数（MFCC）特征,使用卷积神经网络（CNN）提取图像特征,然后利用典型相关分析将两种特征映射到子空间并用欧氏距离进行检索,并在慕尼黑工业大学触觉纹理数据集上进行实验验证,实现了使用声音检索图像的跨模态检索过程。实验结果表明,所提出的方法在材质检索方面有较好应用效果。     

基于特征融合的海洋哺乳动物声音识别

为了提高海洋哺乳动物声音识别算法的识别率和鲁棒性,提出了一种将梅尔倒谱系数MFCC、线性倒谱系数LFCC和时域特征融合作为特征参数进行声音识别的方法。该方法通过融合不同倒谱系数以增强对不同频段的表征能力,通过融合时域特征来更全面地描述声音信息。声音样本通过基于海洋环境下的预处理、特征提取与融合后,用支持向量机进行分类识别。相对于传统算法只针对一种或几种哺乳动物进行识别,该方法在包含61种海洋哺乳动物声音的样本库中进行测试。测试结果显示该算法较传统的梅尔倒谱系数在识别率上提升了5.5%,且在海洋低信噪比环境下有更好的识别表现。     

基于计算机辅助汉语教学系统的发音评分

提出了一种适用于母语非汉语学习者（Learner with Mandarin as a second language,ML2）学习汉语的自动发音评分（Automatic Pronunciation Evaluation,APE）方法。引入双语料语音库,将发音评分分为发音方式和发音效果两部分评分,提取各部分相应的特征参数,提出一种 双语音动态时间规整（Bilingual Dynamic Time Warping,BDTW）的方法来匹配比较得到语音相似度,通过评分机制得到最后的评分结果。通过实验验证了该方法的有效性。     

基于VQ的说话人自动识别系统的实现

文中以语音信号的LPC倒谱系数、△倒谱系数、基音周期和△基音周期的混合特征参数作为识别说话人的特征，运用VQ技术实现了说话人自动识别。在10个说话人，1800个汉语数字和单词语音的语音库上进行了系统的识别实验，其中单音节语音的平均识别率达到了92％，双音节语音达到了96．67％，四音节语音达到了97．67％。     

基于LSTM神经网络的油浸式变压器异常声纹诊断方法研究

利用声音信号对电力变压器进行状态诊断是一种不停机、无接触的设备维护方法,可以诊断变压器异常状态类型。提出了一种基于LSTM神经网络的电力变压器异常诊断的方法,采集电力变压器在正常状态、过载和放电3种运行状态下发出的声音信号,将声音信号进行预处理并提取声音信号的MFCC特征,再将其通过一、二阶差分组合成一组声音特征的矢量,输入LSTM神经网络中进行训练。训练结果表明,将LSTM神经网络应用在电力变压器状态声音诊断上对3种状态的识别均能达到99%以上的准确率。     

基于高斯滤波器及费舍尔准则的特征提取方法

针对梅尔倒谱系数与翻转梅尔倒谱系数在语种识别应用中的不足,采用高斯滤波器代替三角滤波器,提出一种新的梅尔倒谱系数提取方法,解决传统梅尔倒谱系数提取中邻近滤波器相关性较弱的问题,并结合Fisher准则构造出最优混合特征参数,采用高斯混合模型分别对不同混合特征进行语种识别。实验结果表明:基于高斯滤波器及Fisher准则的改进梅尔倒谱系数混合特征参数作为语种识别特征具有较高的识别准确率。