共查询到20条相似文献,搜索用时 93 毫秒
1.
2.
3.
4.
针对支持向量机(Support Vector Machine,SVM)的参数优化问题,提出了一种改进的混合蛙跳算法(Improved Shuffled Frog Leaping Algorithm,Im-SFLA),提高了其在实用语音情感识别中的学习能力。首先,我们在SFLA中引入了模拟退火(Simulated Annealing,SA)、免疫接种(Immune Vaccination,IV)、高斯变异和混沌扰动算子,平衡了搜索的高效性和种群的多样性;第二,利用Im-SFLA优化SVM的参数,提出了一种Im-SFLA-SVM方法;第三,分析了烦躁等实用语音情感的声学特征,重点分析了基音、短时能量、共振峰和混沌特征随情感类别的变化特性,构建出144维的情感特征向量并采用LDA降维到4维;最后,在实用语音情感数据库上测试了算法性能,将提出的算法与混合蛙跳算法(Shuffled Frog Leaping Algorithm,SFLA)优化SVM参数的方法(SFLA-SVM方法)、粒子群优化(Particle Swarm Optimization,PSO)算法优化SVM参数的方法(PSO-SVM方法)、基本SVM方法、高斯混合模型(Gaussian Mixture Model,GMM)方法和反向传播(Back Propagation,BP)神经网络法等进行对比。实验结果表明,采用Im-SFLA-SVM方法的平均识别率达到77.8%,分别高于SFLA-SVM方法、PSO-SVM方法、SVM方法、GMM方法和BP神经网络法各1.7%,2.7%,3.4%,4.7%,7.8%,并且对于烦躁这种实用情感的识别率提高效果最为明显,从而证实了Im-SFLA是一种有效的SVM参数选择方法,并且Im-SFLA-SVM方法能显著提升实用语音情感的识别率。 相似文献
5.
6.
思想和方法是高中物理解题的关键。学生只有具备了解题思想,掌握了相应的解题方法,才能够快速准确地解决物理习题。高中物理解题的思想、方法主要有图像法、模型法、等效法、整体法与隔离法、守恒法、极值法等,对这些方法进行深入的探究,具有现实的理论意义和实践价值。 相似文献
7.
激光三角法具有非接触测量、测量范围大、相对测量精度高、结构简单、环境适应性强等多种优点, 得到了广泛应用。但是三角测量的理论公式具有非线性特征, 而且光学结构参数(a、b、θ)等在现实工程中具有不可测性。研究了三角测量中数学模型的建立方法, 选用多项式展开方法建立数学模型。通过应用最小二乘法拟合多项式的方法求解模型系数, 提出了根据最大相对拟合残差要求、结合相关系数用于控制拟合多项式阶数的评价方法, 并通过实际光学系统验证了该方法的可行性, 达到了0.01%的相对误差。最小二乘法拟合多项式的方法对于激光三角位移传感器的标定和系统误差消除具有实际的指导意义。 相似文献
8.
9.
10.
超声弯曲模式变幅杆的振动分析 总被引:5,自引:0,他引:5
介绍了用传递矩阵分析超声弯曲变幅杆的方法,推导了变幅杆的频率方程、放大系数、形状因子的表达式。作为方法的应用例,对圆锥形变幅杆进行了一些数值计算。最后进行了验证。结果表明传递矩阵法是弯曲变幅杆分析的有效方法。 相似文献
11.
12.
The spectral resolution of the binaural system was measured using a tone-detection task in a binaural analog of the notched-noise technique. Three listeners performed 2-interval, 2-alternative, forced choice tasks with a 500-ms out-of-phase signal within 500 ms of broadband masking noise consisting of an "outer" band of either interaurally uncorrelated or anticorrelated noise, and an "inner" band of interaurally correlated noise. Three signal frequencies were tested (250, 500, and 750 Hz), and the asymmetry of the filter was measured by keeping the signal at a constant frequency and moving the correlated noise band relative to the signal. Thresholds were taken for bandwidths of correlated noise ranging from 0 to 400 Hz. The equivalent rectangular bandwidth of the binaural filter was found to increase with signal frequency, and estimates tended to be larger than monaural bandwidths measured for the same listeners using equivalent techniques. 相似文献
13.
R P Carlyon 《The Journal of the Acoustical Society of America》1987,81(2):418-426
Thresholds for 10-ms sinusoids simultaneously masked by bursts of bandpass noise centered on the signal frequency were measured for a wide range of signal frequencies and noise levels. Thresholds were defined as the signal power relative to the masker power at the output of an auditory filter centered on the signal frequency. It was found that the presentation of a continuous random noise, with a spectral notch centered on the signal frequency, produced a reduction in signal thresholds of up to 11 dB. A notched noise spectrum level of 0-5 dB above that of the masker proved most effective in producing a masking release, as measured by a reduction in masked threshold. A release from masking of up to 7 dB could be obtained with a continuous bandpass noise. The most effective spectrum level of this noise was 5 dB below that of the masker. The effect of the continuous notched noise was to reduce signal-to-masker ratios at threshold to about 0 dB, regardless of the threshold in the absence of continuous noise. Thus the greatest release from masking occurred when "unreleased" thresholds were highest. The release from masking is almost complete within 320 ms of notched noise onset, and persists for about 160 ms after notched noise offset, regardless of notched noise level. The phenomenon is similar in many ways to the "overshoot" effect reported by Zwicker [J. Acoust. Soc. Am. 37, 653-663 (1965)]. It is argued that both effects can be largely attributed to peripheral short-term adaptation, a mechanism which is also believed to be involved in forward masking. 相似文献
14.
J W Hall 《The Journal of the Acoustical Society of America》1986,79(3):781-787
Masking noise well separated in frequency from the signal may improve the detectability of the signal if the masking noise is modulated. This effect is referred to as co-modulation masking release (CMR). The present experiments examine the effect of across-frequency differences in masking noise level on CMR. Three experiments were performed, each using a different method to create modulated noise stimuli having across-frequency differences in the spectrum level. All stimulation was monaural. Experiment I used a notched noise method (selectively reducing the level for the critical band centered on the signal). Experiment II used a method in which the level of a 100-Hz-wide masker centered on the signal was varied, and flanking noise bands were of constant level. Experiment III used a method in which flanking noise bands were varied in level, and the 100-Hz-wide masker centered on the signal was of constant level. The signal was a 1000-Hz, 300-ms pure tone. The CMR effect was negated by small spectral notches centered on the signal (experiment I). However, CMR proved to be relatively robust to across-frequency level differences in experiments II and III (a CMR effect occurred for across-frequency differences in spectrum level as great as 20 dB). Low CMR's obtained in experiment I were probably due to relatively poor correlation of across-frequency modulation pattern which occurred with notched noise. The results of experiments II and III suggest that the fluctuation pattern is of primary importance in providing release from masking, and that information on absolute levels, coded across frequency, is of less importance. 相似文献
15.
Thresholds of a 5-ms, 1-kHz signal were determined in the presence of a frozen-noise masker. The noise had a flat power spectrum between 20 Hz and 5 kHz and was presented with a duration of 300 ms. The following interaural conditions were tested with four listeners: Noise and signal monaural at the same ear (monaural condition, NmSm), noise and signal identical at both ears (diotic condition, NoSo), noise identical at both ears and signal monaural (dichotic condition, NoSm) and uncorrelated noise at the two ears and signal monaural (NuSm). The signal was presented at a fixed temporal position with respect to the frozen noise in all measurements and thresholds were determined for different starting phases of the carrier frequency of the signal. Variation of the carrier phase strongly influenced the detection in the diotic condition and the masked thresholds varied by more than 10 dB. The pattern of thresholds for the monaural condition was less variable and the thresholds were generally higher than for the diotic condition. The monaural-diotic difference for specific starting phases amounted to as much as 8 dB. Comparison measurements using running noise maskers revealed no such difference. This relation between monaural and diotic thresholds was further investigated with eight additional subjects. Again, monaural and diotic thresholds in running noise were identical, while in frozen noise, diotic thresholds were consistently lower than monaural thresholds, even when the ear with the lower NmSm threshold was compared. For the starting phase showing the largest monaural-diotic difference, the thresholds for NoSm lay between the monaural and the diotic values. At other starting phases, the NoSm threshold was clearly lower than both the NmSm and the NoSo threshold. One possible explanation of the observed monaural-diotic differences relates to contralateral efferent interaction between the right and the left hearing pathway. A prediction based on this explanation was verified in a final experiment, where frozen-noise performance for NmSm was improved by simultaneously presenting an uncorrelated running noise to the opposite ear. 相似文献
16.
The objectives of this study were to measure suppression with bandlimited noise extended below and above the signal, at lower and higher signal frequencies, between younger and older subjects, and between subjects with normal hearing and cochlear hearing loss. Psychophysical suppression was assessed by measuring forward-masked thresholds at 0.8 and 2.0 kHz in bandlimited maskers as a function of masker bandwidth. Bandpass-masker bandwidth was increased by introducing noise components below and above the signal frequency while keeping the noise centered on the signal frequency, and also by adding noise below the signal only, and above the signal only. Subjects were younger and older adults with normal hearing and older adults with cochlear hearing loss. For all subjects, suppression was larger when noise was added below the signal than when noise was added above the signal, consistent with some physiological evidence of stronger suppression below a fiber's characteristic frequency than above. For subjects with normal hearing, suppression was greater at higher than at lower frequencies. For older subjects with hearing loss, suppression was reduced to a greater extent above the signal than below and where thresholds were elevated. Suppression for older subjects with normal hearing was poorer than would be predicted from their absolute thresholds, suggesting that age may have contributed to reduced suppression or that suppression was sensitive to changes in cochlear function that did not result in significant threshold elevation. 相似文献
17.
18.
A linear system driven by correlated asymmetric dichotomous noises and periodic signal was investigated in the overdamped
case. The exact expressions of output signal amplitude and signal-to-noise ratio (SNR) of the system were derived. By means
of numerical calculations, we found that: (i) At some fixed multiplicative noise intensities, the output signal amplitude
with frequency exhibits the structure of a weak peak, even no peak as the dichotomous noise is asymmetric; (ii) In the case
of asymmetric dichotomous noise, the signal frequency can cause non-monotonous behavior of the output signal amplitude with
respect to multiplicative noise intensity; (iii) The curve of SNR with frequency has a weak peak and a trough in the case
of symmetric dichotomous noise, but no peak with asymmetric; (iv) Whether the multiplicative noise is symmetric or asymmetric,
the noise can enhance response of the system; (v) The SNR increases with the correlation strength between the two noises decreasing.
In addition, the plane of multiplicative noise intensity versus noise symmetric parameter was plotted. 相似文献
19.
Signal detection was determined in conditions where the masker was a 10-Hz-wide noise band centered on the signal, and in conditions where either a comodulated or noncomodulated noise band (centered at 0.8 times the signal frequency) was also present. Signal frequencies of 500 or 2000 Hz were investigated. In one condition of the first experiment, the signal was exactly the same 10-Hz-wide noise band as the masker, added to the masker in phase. This condition was designed to limit the availability of cues based upon dip listening, suppression, beating, or across-frequency differences in noise envelope correlation, but to afford a cue based upon across-frequency envelope amplitude difference. The narrow-band noise signal resulted in approximately the same magnitude of comodulated masking release (CMR) as was found for a pure-tone signal. This result suggested that one important cue for CMR is an across-frequency difference in envelope amplitude. Stimulus conditions in the second experiment were intended to disrupt cues of across-frequency envelope amplitude difference, but to afford cues based upon across-frequency differences in noise envelope correlation. In this experiment, cues based upon envelope amplitude were impoverished by randomly varying the level of the flanking band from interval to interval, and by adjusting the level in the on-signal band to be the same in the nonsignal intervals as the level of noise plus signal in the signal interval. Again, substantial CMRs occurred, suggesting that another cue for CMR may be envelope pattern or correlation. The results of these experiments indicated that CMR is probably based upon more than one stimulus variable. 相似文献
20.
Modulation detection thresholds (as a function of sinusoidal amplitude modulation frequency) and temporal gap detection thresholds were measured for three low-pass-filtered noise signals (fc = 1000, 2000, and 4000 Hz), a high-pass-filtered noise signal (fc = 4000 Hz), and a broadband signal. The two latter noise signals were effectively low-pass filtered (fc = 6500 Hz) by the earphone. Each of the filtered signals was presented with a complementary filtered noise masker. Modulation and gap detection thresholds were lowest for the broadband and high-pass signals. Thresholds were significantly higher for the low-pass signals than for the broadband and high-pass signals. For these tasks and conditions, the high-frequency content of the noise signal was more important than was the signal bandwidth. Sensitivity (s) and time constant (tau) indices were derived from functions fitted to the modulation detection data. These indices were compared with gap detection thresholds for corresponding signals. The gap detection thresholds were correlated inversely (rho = -1.0, p less than 0.05) with s (i.e., smaller gap detection thresholds were correlated with greater sensitivity to modulation), but were not correlated significantly with tau, which was relatively invariant across signal conditions. 相似文献