提高汉语清晰度的语音处理

根据语音不同频区在清晰度中所起的作用及感音性聋患者听力改变的多样性,作者设计和试制了一台具有滤波、动态范围压缩及频率补偿等功能的TB—1型听力补偿测定仪。用它可以选定不同患者所需的最佳听力补偿方式,为配制专用助听器提供依据。选择该仪器提供的语音处理,可使正常人在信噪比为odB时的语音清晰度从66.6％提高到87.4％;使感音性聋患者在安静环境、信噪比为6dB和odB时的语音清晰度分别从42.1％提高到66.0％,从45.8％提高到78.0％及从40.3％提高到71.0％。文章讨论了该仪器的语音处理方式及其应用方面的一些问题。     

母语为汉语的听者听英语时的空间去掩蔽现象研究

通过心理声学实验研究了来自不同方向具有不同信噪比的两种十扰声条件下,母语为汉语的听者对英语的空间去掩蔽现象.在消声室指定位置布放扬声器,发出目标声和干扰声,通过听者对目标卢进行听音识别,得到听者识别的正确率.实验结果显示:只在正前方播放目标语音时,识别正确率大于99%,当目标和干扰语音都位于听者正前方时,正确率为57%;当目标和干扰语音随机位于±60°时,正确率为96%;特别地,当甘标语音和干扰信号都位于听者正前方时,若干扰为噪声,随着信噪比从0 dB降低到-12 dB,正确率从96%降低到34%,而当干扰为语音时,随着信噪比从0 dB降低到-12 dB,正确率先足下降,随后有平均幅度为27%的明显上升,在此之后又是下降的趋势;当噪声干扰和语音干扰位于60°时,随着信噪比从-4 dB降低到-16 dB,正确率分别从99%降低到80%和从98%降低到91%.研究表明:空间分离对于母语为汉语的听者的英语语音町懂度有明显增益;大多数情况下英语语音的正确率都随着信噪比的降低而下降.这和对母语为其他语言的相关研究结论一致.     

基于光子相关光谱分析的声压量值复现方法

寻求不依赖于实验室标准传声器的灵敏度而直接溯源至国际单位制基本单位的声压量值复现技术是声学计量的长期目标,对声压量值摆脱实物基准具有重要意义,激光多普勒测速技术是实现这一目标的有效途径.以行波管内平面波声场为测量对象,建立无固定频移的激光多普勒测速系统,采用光子相关光谱分析法解调多普勒信号,获得声管内示踪粒子的振动速度,根据平面波声压与质点振动速度的线性关系,复现声管测量点处的声压.以工作标准传声器的测量结果为参考,评估测量方案的可行性和测量结果的准确性,分析影响测量准确性的主要因素.测量结果表明,声波频率为315Hz,声压级在100dB~110dB范围内间隔1dB变化时,测量偏差小于0.5dB;声压级为105dB,声波频率为315,400,500,800Hz时测量偏差小于0.3dB.     

语音特性及声场因素对老年人警报语音可懂度和主观感受的影响*

警报语音广播是紧急情况时提高建筑物内疏散效率的有效手段。通过实验室研究的方法,研究了语音特性及声场因素对老年群体警报语音可懂度和主观感受的影响。主观感受选取了听音容易程度和感知紧迫性两个维度。研究结果表明,老年人警报语音可懂度和听音容易程度评价主要受语速、信噪比和混响时间的影响,且呈现一致的变化趋势,即随着语速和混响时间降低以及信噪比的增加,老年人可懂度和听音容易程度得分均升高,而声压级(最低设置为60dB)、有无警铃和噪声类型没有显著影响。感知紧迫性随语速和声压级的增加而显著增加,信噪比、混响时间及有无警铃声对感知紧迫性并无显著影响。采用人声播报的警报语音其可懂度和感知紧迫性显著高于合成声。比较老年人和年轻人群体的结果发现,在语速、声压级和噪声类型对主观评价的影响上有显著差异。为建立老年人理想且安全的声环境,应采用人声播报并适当降低语速以保证可懂度,同时混响及信噪比条件两方面的改善都是必要的。     

小波包自适应阈值语音降噪新算法

为了克服低信噪比输入下,语音增强造成语音清音中的弱分量损失,造成重构信号包络失真的问题。论文提出了一种新的语音增强方法。该方法根据语音感知模型,采用不完全小波包分解拟合语音临界频带,并对语音按子带能量进行清浊音区分处理,在阈值计算上,提出了一种清浊音分离,基于子带信号能量的小波包自适应阈值算法。通过仿真实验,客观评测和听音测试表明,该算法在低信噪比输入时较传统算法,能够更加有效地减少重构信号包络失真,在不损伤语音清晰度和自然度的前提下,使输出信噪比明显提高。将该算法与能量谱减法结合,进行二次增强能进一步提高降噪输出的语音质量。     

基于悬臂梁调谐技术的光纤光栅无源振动监测

采用匹配光纤光栅设计了一种结构简单的振动信号无源监测装置.该装置利用悬臂梁调谐技术能够将微小振动信号转化为光电探测器可探测的光强信号,利用示波器实现实时监测.实验中对振幅为3mm的简谐振动信号进行了监测,测量结果与振动频率一致,可测量7～20Hz的振动,信噪比不低于14.9dB.监测频率受限是因为悬臂梁的性质,如采用金属材料或者采用齿轮组对转子进行减速,该装置可探测更高的频率.     

仰卧人体在三轴单向全身振动激励下的振动传递特性

研究了人体仰卧姿态下,振动的幅值大小和输入方向对人体头部和胸部振动传递率的影响。分别采用频率范围为0.5～20 Hz,振动幅值为0.2～0.8 m/s² r.m.s.的纵向、横向和垂直单轴随机振动激励,测量了12名男性被试者自振源至头部和胸部的振动传递特性。结果表明,人体头部和胸部振动传递率的共振频率在纵向方向上大约为4 Hz,在横向方向上大约为2.4 Hz,而在垂直方向上头部和胸部传递率的共振频率并不相同。在纵向激励下,仰卧人体各部位振动传递率随振幅变化而表现出的非线性特征最为明显。     

带加速度补偿的DFB光纤激光水听器

针对分布反馈式光纤激光水听器在用于水声探测时极易受加速度效应干扰的问题,设计了一种双膜片对称结构的光纤激光水听器,对该水听器的抗加速度性能进行了研究.建立了双膜片结构水听器的加速度灵敏度理论模型,分析了水听器各部件的尺寸大小、材料参量与水听器加速度灵敏度的关系,实现了对水听器结构的优化设计;加工制作了分布反馈式光纤激光水听器原型样品,并进行了实验研究.测量结果表明,在2.5~10kHz的频率范围内,该结构水听器的平均声压灵敏度为-132.6dB,波动幅度不大于±0.5dB,加速度灵敏度小于-28dB.该水听器在保证了较高声压灵敏度与平坦的响应曲线的同时,抗加速度性能也得了有效改善,可大大提高光纤激光水听器阵列在运动状态下对远距离目标探测的信噪比.     

复合听神经动作电位与调幅刺激声包络的时序相关分析

在豚鼠用植入内听道出口处的电极记录了调幅声诱发的复合听神经动作电位，并研究了在声参数系统地变化时反应与调制波之间的时序相关性。所用的调制波有三种：频率固定的连续或短段正弦波（频率范围40Hz－5kHz），频率变化的短段正弦波及小段语音信号。其它声参数变化范围为：载波500Hz－20kHz，调制深度5％－95％，强度20－90dBSPL。对于频率固定的连续或短段正弦调制波，在大多数参数条件下，反应与调制波之间的相关系数（r）相当高：0．80－0．95，在由于参数不适当、反应幅度下降时，r相应地变小。对于频率变化的短段正弦调制波，r在0．66－0．86之间变化。当用语音信号片段作调制波时，反应与调制波之间仍存在一定的相关性（r在0．50左右变化），表明对语音信息在耳蜗水平的编码，时间模式是有效的。文中对听觉时间机理的一些理论及技术要点作了讨论。     

正弦振动对航空相机成像影响的分析与验证

航空相机拍照时,由于受到振动影响成像分辨率会降低。采用动态调制传递函数研究正弦振动对成像的影响,通过基于空间域的计算方法对高频和低频振动的调制传递函数(MTF)进行了分析,指出了曝光时间不等于振动周期整数倍的一般情况时,高频振动对成像质量的影响也具有随机性,但没有低频振动的随机性明显。采用快速反射镜作为振动源进行成像实验,对获得的振动图像使用小波变换提取图像的高频信息。实验中将曝光时间设定为20ms,振动频率为50Hz时,高频信息波动范围仅为0.0582×104,可以认为基本没有波动。当振动频率为15Hz时,波动范围为0.6233×104,随机性很明显,而振动频率为65Hz时,波动范围为0.1245×104,随机性不明显。通过高频信息与MTF的对应关系,实验证明了理论分析的正确性。在理论分析和实验的基础上,总结了正弦振动对成像影响的普遍性结论。该结论可用于分析成像系统光机结构设计,也可应用于图像去模糊,具有一定的意义。     

Factors affecting perception thresholds of vertical whole-body vibration in recumbent subjects: Gender and age of subjects, and vibration duration

Some factors that may affect human perception thresholds of the vertical whole-body vibrations were investigated in two laboratory experiments with recumbent subjects. In the first experiment, the effects of gender and age of subjects on perception were investigated with three groups of 12 subjects, i.e., young males, young females and old males. For continuous sinusoidal vibrations at 2, 4, 8, 16, 31.5 and 63 Hz, there were no significant differences in the perception thresholds between male and female subjects, while the thresholds of young subjects tended to be significantly lower than the thresholds of old subjects. In the second experiment, the effect of vibration duration was investigated by using sinusoidal vibrations, at the same frequencies as above, modulated by the Hanning windows with different lengths (i.e., 0.5, 1.0, 2.0 and 4.0 s) for 12 subjects. It was found that the peak acceleration at the threshold tended to decrease with increasing duration of vibration. The perception thresholds were also evaluated by the running root-mean-square (rms) acceleration and the fourth power acceleration method defined in the current standards. The differences in the threshold of the transient vibrations for different durations were less with the fourth power acceleration method. Additionally, the effect of the integration time on the threshold was investigated for the running rms acceleration and the fourth power acceleration. It was found that the integration time that yielded less differences in the threshold of vibrations for different durations depended on the frequency of vibration.     

Development of noise and vibration ride comfort criteria.

A laboratory investigation was directed at the development of criteria for the prediction of ride quality in a noise-vibration environment. The stimuli for the study consisted of octave bands of noise centered at 500 and 2000 Hz and vertical floor vibrations composed of either 5 Hz sinusoidal vibration, or random vibrations centered at 5 Hz and with a 5 Hz bandwidth. The noise stimuli were presented at A-weighted sound pressure levels ranging from ambient to 95 dB and the vibration and acceleration levels ranging from 0.02--0.13 grms. Results indicated that the total subjective discomfort response could be divided into two subjective components. One component consisted of subjective discomfort to vibration and was found to be a linear function of vibration acceleration level. The other component consisted of discomfort due to noise which varied logarithmically with noise level (power relationship). However, the magnitude of the noise discomfort component was dependent upon the level of vibration present in the combined environment. Based on the experimental results, a model of subjective discomfort that accounted for the interdependence of noise and vibration was developed. The model was then used to develop a set of criteria (constant discomfort) curves that illustrate the basic design tradeoffs available between noise and vibration.     

EFFECT OF PHASE ON HUMAN RESPONSES TO VERTICAL WHOLE-BODY VIBRATION AND SHOCK—ANALYTICAL INVESTIGATION

The effect of the “phase” on human responses to vertical whole-body vibration and shock has been investigated analytically using alternative methods of predicting subjective responses (using r.m.s., VDV and various frequency weightings). Two types of phase have been investigated: the effect of the relative phase between two frequency components in the input stimulus, and the phase response of the human body. Continuous vibrations and shocks, based on half-sine and one-and-a-half-sine accelerations, each of which had two frequency components, were used as input stimuli. For the continuous vibrations, an effect of relative phase was found for the vibration dose value (VDV) when the ratio between two frequency components was three: about 12% variation in the VDV of the unweighted acceleration was possible by changing the relative phase. The effect of the phase response of the body represented by frequency weightings was most significant when the frequencies of two sinusoidal components were about 3 and 9 Hz. With shocks, the effect of relative phase was observed for all stimuli used. The variation in the r.m.s. acceleration and in the VDV caused by variations in the relative phase varied between 3 and 100%, depending on the nature of stimulus and the frequency weighting. The phase of the frequency weightings had a different effect on the r.m.s. and the VDV.     

Equal sensation curves for whole-body vibration expressed as a function of driving force

Previous studies have shown that the seated human is most sensitive to whole-body vertical vibration at about 5 Hz. Similarly, the body shows an apparent mass resonance at about 5 Hz. Considering these similarities between the biomechanical and subjective responses, it was hypothesized that, at low frequencies, subjective ratings of whole-body vibration might be directly proportional to the driving force. Twelve male subjects participated in a laboratory experiment where subjects sat on a rigid seat mounted on a shaker. The magnitude of a test stimulus was adjusted such that the subjective intensity could be matched to a reference stimulus, using a modified Bruceton test protocol. The sinusoidal reference stimulus was 8-Hz vibration with a magnitude of 0.5 m/s2 rms (or 0.25 m/s2 rms for the 1-Hz test); the sinusoidal test stimuli had frequencies of 1, 2, 4, 16, and 32 Hz. Equal sensation contours in terms of seat acceleration showed data similar to those in the literature. Equal sensation contours in terms of force showed a nominally linear response at 1, 2, and 4 Hz, but an increasing sensitivity at higher frequencies. This is in agreement with a model derived from published subjective and objective fitted data.     

Factors affecting the loudness of modulated sounds.

Loudness matches were obtained between unmodulated carriers and carriers that were amplitude modulated either periodically (rates between 2 and 32 Hz, modulation sinusoidal either on a linear amplitude scale or on a dB scale; the latter is called dB modulation) or with the envelope of the speech of a single talker. The carrier was a 4-kHz sinusoid, white noise, or speech-shaped noise. Both normally hearing subjects and subjects with cochlear hearing loss were tested. Results were expressed as the root-mean-square (rms) level of the modulated carrier minus the level of the unmodulated carrier at the point of equal loudness. If this difference is positive, this indicates that the modulated carrier has a higher rms level at the point of equal loudness. For normally hearing subjects, the results show: (1) For a 4000-Hz sinusoidal carrier, the difference was slightly positive (averaging about 0.7 dB). There was no significant effect of modulation rate or level over the range 20-80 dB SL. (2) For a speech-shaped noise or white noise carrier, the difference was close to zero, although for large modulation depths it tended to be negative. There was no clear effect of level (over the range 35-75 dB SPL) or modulation rate. For the hearing-impaired subjects, the differences were small, but tended to be slightly negative for both the 4000-Hz carrier and the noise carriers, when the modulation rate was above 2 Hz. Again, there was no clear effect of overall level. However, for dB modulation, the differences became more negative with increasing modulation depth. For modulation rates in the range 4-32 Hz, the results could be fitted reasonably well using the assumption that the loudness of modulated sounds is based on the rms value of the time-varying intensity of the response of the basilar membrane (taking into account the compression that occurs in the normal cochlea). The implications of the results for the fitting of multi-band compression hearing aids and for the design of loudness meters are discussed.     

Effect of phase on discomfort caused by vertical whole-body vibration and shock--experimental investigation

An experimental study has investigated the effect of "phase" on the subjective responses of human subjects exposed to vertical whole-body vibration and shock. The stimuli were formed from two frequency components: 3 and 9 Hz for continuous vibrations and 3 and 12 Hz for shocks. The two frequency components, each having 1.0 ms(-2) peak acceleration, were combined to form various waveforms. The effects of the vibration magnitude on the discomfort caused by the input stimuli were also investigated with both the continuous vibrations and the shocks. Various objective measurements of acceleration and force at the seat surface, the effects of different frequency weightings and second and fourth power evaluations were compared with judgments of the discomfort of the stimuli. It was found that a 6% to 12% increase in magnitude produced a statistically significant increase in discomfort with both the continuous vibrations and the shocks. Judgments of discomfort caused by changes in vibration magnitude were highly correlated with all of the objective measurements used in the study. The effects on discomfort of the phase between components in the continuous vibrations were not statistically significant, as predicted using evaluation methods with a power of 2. However, small changes in discomfort were correlated with the vibration dose value (VDV) of the Wb frequency-weighted acceleration. The effect of phase between frequency components within the shocks was statistically significant, although no objective measurement method used in the study was correlated with the subjective judgments.     

EFFECT OF MUSCLE TENSION ON NON-LINEARITIES IN THE APPARENT MASSES OF SEATED SUBJECTS EXPOSED TO VERTICAL WHOLE-BODY VIBRATION

In subjects exposed to whole-body vibration, the cause of non-linear dynamic characteristics with changes in vibration magnitude is not understood. The effect of muscle tension on the non-linearity in apparent mass has been investigated in this study. Eight seated male subjects were exposed to random and sinusoidal vertical vibration at five magnitudes (0·35-1·4 m/s² r.m.s.). The random vibration was presented for 60 s over the frequency range 2·0-20 Hz; the sinusoidal vibration was presented for 10 s at five frequencies (3·15, 4·0, 5·0, 6·3 and 8·0 Hz). Three sitting conditions were adopted such that, in two conditions, muscle tension in the buttocks and the abdomen was controlled. It was assumed that, in these two conditions, involuntary changes in muscle tension would be minimized. The force and acceleration at the seat surface were used to obtain apparent masses of subjects. With both sinusoidal and random vibration, there was statistical support for the hypothesis that non-linear characteristics were less clear when muscle tension in the buttocks and the abdomen was controlled. With increases in the magnitude of random vibration from 0·35 to 1·4 m/s² r.m.s., the apparent mass resonance frequency decreased from 5·25 to 4·25 Hz with normal muscle tension, from 5·0 to 4·38 Hz with the buttocks muscles tensed, and from 5·13 to 4·5 Hz with the abdominal muscles tensed. Involuntary changes in muscle tension during whole-body vibration may be partly responsible for non-linear biodynamic responses.     

The vibration discomfort of standing persons: 0.5-16-Hz fore-and-aft, lateral, and vertical vibration

To minimise the discomfort of standing people caused by vibration of a floor, it is necessary to know how their sensitivity to vibration depends on the frequency of the vibration. This study was designed to determine how the discomfort of standing people exposed to horizontal and vertical vibration depends on vibration frequency over the range 0.5-16 Hz. Using the method of magnitude estimation, sixteen subjects judged the discomfort caused by fore-and-aft, lateral, and vertical sinusoidal vibration at each of the sixteen preferred one-third octave centre frequencies from 0.5 to 16 Hz at each of nine magnitudes. Subjects also reported the main cause of their discomfort. Equivalent comfort contours were constructed, reflecting the effect of frequency on subject sensitivity to vibration acceleration. With horizontal vibration, at frequencies between 0.5 and 3.15 Hz the discomfort was similar when the vibration velocity was similar, whereas at frequencies between 3.15 and 16 Hz the discomfort was similar when the vibration acceleration was similar. At frequencies less than 3.15 Hz, the subjects experienced problems with their stability, whereas at higher frequencies vibration discomfort was mostly experienced from sensations in the legs and feet. With vertical vibration, discomfort was felt in the lower-body and upper-body at all frequencies. The frequency weightings in current standards for predicting the vibration discomfort of standing persons have been greatly influenced by the findings of studies with seated subjects: the weightings are consistent with the experimentally determined frequency-dependence of discomfort caused by vertical vibration but inconsistent with the experimentally determined frequency-dependence of discomfort caused by horizontal vibration. The results suggest that the responses of seated and standing people are similar for vertical vibration, but differ for horizontal vibration, partly due to greater instability in standing persons.     

Middle-ear response in the chinchilla and its relationship to mechanics at the base of the cochlea

The responses of the malleus and the stapes to sinusoidal acoustic stimulation have been measured in the middle ears of anesthetized chinchillas using the M?ssbauer technique. With "intact" bullas (i.e., closed except for venting via capillary tubing), the vibrations of the tip of the malleus reach a maximal peak velocity of about 2 mm/s in responses to 100-dB SPL tones in the frequency range 500-6000 Hz; vibration velocity diminishes toward lower frequencies with a slope of about 6 dB/oct. Opening the bulla widely increases the responses to low-frequency stimuli by as much as 16 dB. At low frequencies, malleus response sensitivity with either open or intact bullas far exceeds all previous measurements in cats and matches or exceeds such measurements in guinea pigs. Whether measured in open or intact bullas, phase-versus-frequency curves closely approximate those predicted from the magnitude-versus-frequency curves by minimum phase theory. The stapes responses are similar to those of the malleus, except that stapes response magnitude is lower, on the average, by 7.5 dB at frequencies below 2 kHz and 10.7 dB at 2 kHz and above. Comparison of the responses of the middle ear with those of the basilar membrane at a site 3.5 mm from the stapes indicates that, at frequencies below 150 Hz, the basilar membrane displacement is proportional to stapes acceleration. At frequencies between 150 and 2000 Hz, basilar membrane displacement is proportional to stapes velocity.     

一种采用双换能器和摆式结构的宽频振动能量采集器

针对悬臂梁振动能量采集器在大振幅振动下梁容易断裂的缺点,本文提出了一种基于摆式结构的具有宽频和倍频特性的振动能量采集器,该采集器由两个Terfenol-D/PMN-PT/Terfenol-D磁电换能器和嵌有六个磁铁的旋转摆构成.文中建立了摆式结构的摆动方程,分析了采集器的频率响应特性以及谐振时的机-磁-电转换特性,并对采集器输出电压波形进行了频谱分析.理论和实验研究表明:该采集器具有宽频和倍频特性,采集器样机在1 g(1 g=9.8 m/s~2)有效值加速度振动下,向下扫频时的半功率带宽达到4.8 Hz,且能在f=16.9 Hz的振动下获得3.569 mW的负载功率.利用双换能器以及采集器的倍频和宽频特性,能有效地提高低频时采集器的输出功率.