Random Noise Technique Research Articles

Enhancing the transferability of adversarial samples with random noise techniques

Deep neural networks have achieved remarkable success in the field of computer vision. However, they are susceptible to adversarial attacks. The transferability of adversarial samples has made practical black-box attacks feasible, underscoring the importance of research on transferability. Existing work indicates that adversarial samples tend to overfit to the source model, getting trapped in local optima, thereby reducing the transferability of adversarial samples. To address this issue, we propose the Random Noise Transfer Attack (RNTA) to search for adversarial samples in a larger data distribution, seeking the global optimum. Specifically, we suggest injecting multiple random noise perturbations into the sample before each iteration of sample optimization, effectively exploring the decision boundary within an extended data distribution space. By aggregating gradients, we identify a better global optimum, mitigating the issue of overfitting to the source model. Through extensive experiments on the large-scale visual classification task on ImageNet, we demonstrate that our method increases the success rate of momentum-based attacks by an average of 20.1%. Furthermore, our approach can be combined with existing attack methods, achieving a success rate of 94.3%, which highlights the insecurity of current models and defense mechanisms.

Online secondary path modeling method with auxiliary noise power scheduling strategy for multi-channel adaptive active noise control system

Accurate model of secondary paths is very crucial for the multi-channel filtered-X least mean square algorithm applied in adaptive active noise control system. The auxiliary random noise technique is popular for online secondary path modeling during adaptive active noise control operation. This paper proposes a simplified variable step-size strategy and an effective auxiliary noise power scheduling strategy for the multi-channel filtered-X least mean square algorithm. Through a defined indirect error signal, the proposed method can guarantee every online secondary path modeling filter has its own exclusive variable step-size strategy to update their coefficients, and every injected noise has its own exclusive scheduling strategy considering all of the corresponding online secondary path modeling filters. The proposed method can improve the adaptive performance and simplifies the complexity of multi-channel adaptive active noise control system. Computer simulations show that the proposed method gives much better noise reduction and secondary path modeling accuracy at a somewhat faster convergence rate comparing with the competing methods.

Multi-channel adaptive active vibration control of piezoelectric smart plate with online secondary path modelling using PZT patches

Abstract Reducing the vibration of smart structures based on adaptive active vibration control (AAVC) has been extensively studied in recent years, because it is a light weight and effective method for reducing low-frequency structural vibrations. The multi-channel AAVC methodology based on the filtered-X least mean square (FXLMS) algorithm is widely implemented in active control applications owing to its self-adjustment ability to adapt to dynamically varying structures. In this paper, a new multi-channel FXLMS with online secondary path modelling (SPM) is designed based on the auxiliary random noise technique. Through a defined indirect error signal, the proposed variable step-size (VSS) strategies can ensure that every online SPM filter and every active control filter have their own exclusive step size to update the coefficients. Moreover, the proposed auxiliary noise power scheduling (ANPS) strategy can ensure that the variation rules of auxiliary noise power applied to different secondary paths are different. A complete multi-channel AAVC real-time experimental system based on NI compact RIO is set up to conduct the experimental investigation. A series of AAVC control experiments on a piezoelectric smart cantilever plate with PZT sensors and actuators are conducted to demonstrate the validity and efficiency of the proposed method. The experimental results show that the vibration of the smart cantilever plate could be effectively attenuated with a high convergence rate. The proposed methodology has an important advantage in applications where active vibration control of piezoelectric smart structures is required.

Design and implementation of multichannel adaptive active vibration control with online secondary path modeling using piezoelectric transducers patches

Adaptive active vibration control (AAVC) is an effective way for reducing structure vibration at low frequency. AAVC methodology is preferred in AVC system due to its self-adjustment ability to adapt to varying dynamics of the structure. The Filtered-X Least Mean Square (FXLMS) algorithm is widely implemented in active control applications. Accurate secondary path models are very crucial for the implementation in multichannel AAVC system based on FXLMS algorithm. The auxiliary random noise technique is often applied to achieve secondary path modeling (SPM) during online operation. This paper proposes a new multichannel AAVC methodology based on online SPM method. The online SPM error is estimated indirectly to reduce the interaction between the AAVC controller and the online SPM filter. A new variable step-size (VSS) strategy for online SPM filter is proposed based on the estimated online SPM error. A simple but effective auxiliary noise power scheduling (ANPS) method is proposed to eliminate the contribution of the auxiliary noise on the residual vibration. A series of multi-channel AAVC experiments on a cantilever epoxy resin plate with PZT sensors and actuators are presented to demonstrate the performance of the proposed methodology. Experiment results indicate that the proposed method provides very good online SPM accuracy, and the vibration of the smart cantilever plate has been effectively attenuated with high convergence rate.

Active noise control with on-line adaptive algorithm in a duct system

In the case of the transfer function for the secondary path is dependent on time, the on-line method which can model it is continuously must be applied to the active noise control technique. And the adaptive random noise technique among the on-line methods is effective in the narrow-band control. In this method, the signal to noise ratio between random noise for modeling and primary noise is low. Therefore, the estimations of transfer function will be prone to inaccuracies and the convergence time will be too long. Such imperfections will have an influence upon the performance of an active noise controller. In this study, t enhance the signal to noise ratio, the on-line method that is combined the conventional adaptive random noise technique and the adaptive line enhancer, is proposed. By using proposed on-line method, a rigorous system identification and control of primary noise have been implemented.

Circumferential and longitudinal viscoelasticity of human iliac arterial segments in vitro

A random noise technique was used to measure the circumferential and longitudinal dynamic elasticity of human common iliac arteries in vitro. For circumferential measurements the frequency ranged from 0.016 to 20 Hz; the phase lag of diameter behind pressure was found to be almost constant (about 5°) and the Young's modulus of elasticity to increase rapidly at first and then more gradually beyond 1–2 Hz. Somewhat similar results were obtained for longitudinal elasticity. Arterial segments were found to be anisotropic when kept at in vivo length and under normal distending pressure.

Sensor response time monitoring using noise analysis

Random noise techniques in nuclear power plants have been developed for system surveillance and for analysis of reactor core dynamics. The noise signals also contain information about sensor dynamics, and this can be extracted using frequency, amplitude and time domain analyses. Even though noise analysis has been used for sensor response time testing in some nuclear power plants, an adequate validation of this method has never been carried out. This paper presents the results of limited work recently performed to examine the validity of the noise analysis for sensor response time testing in nuclear power plants. The conclusion is that noise analysis has the potential for detecting gross changes in sensor response but it cannot be used for reliable measurement of response time until more laboratory and field experience is accumulated. The method is more advantageous for testing pressure sensors than it is for temperature sensors. This is because: 1) for temperature sensors, a method called Loop Current Step Response test is available which is quantitatively more exact than noise analysis, 2) no method currently exists for on-line testing of pressure transmitters other than the Power-Interrupt test which is applicable only to force balance pressure transmitters, and 3) pressure sensor response time is affected by sensing line degradation which is inherently taken into account by testing with noise analysis.

Calibration of acoustic-emission receivers by a random-noise technique

Calibration of acoustic-emission receivers by a random-noise technique

Frequency dependence of effective nasal resistance.

Forced random noise techniques were used to characterize the effective resistance of the nasal passages in normal subjects and subjects who were candidates for surgical correction of nasal obstruction. The slope of the effective resistance curve was characterized by the average resistances over 3 to 5 Hz and 13 to 15 Hz. Subjects with nasal obstruction had effective nasal resistance curves which were more frequency-dependent than those of normal subjects (p less than 0.05). A mathematical model consisting of two series resistance-inertance elements in parallel was used to simulate the nasal passages. With this model we were able to generate effective resistance curves similar to those measured in both normal and obstructed subjects. The forced random noise technique may be a convenient, noninvasive technique for studying the fluid mechanical properties of airflow in the nasal passages and the effects of pharmacological and surgical interventions on nasal resistance.

Forced noise mechanical parameters during inspiration and expiration

Using the forced random noise technique we collected three 5-min records of pressure and flow signals in five clinically normal nonsmoking adults. Each recording was analyzed to obtain respiratory impedance spectra from data obtained during continuous sampling, midinspiration, and midexpiration. Estimates of respiratory resistance, inertance, and compliance were obtained by regression analysis on each impedance spectrum. Mean expiratory resistance and compliance were significantly (P less than 0.01) greater than corresponding mean inspiratory values, while mean expiratory inertance was significantly (P less than 0.001) less than the mean inspiratory value. Parameters from continuous sampling agreed fairly well with inspiratory resistance and inertance and with expiratory compliance. The coefficient of variation of parameters from continuous data were similar to, or smaller than, those from inspiratory or expiratory data. Thus continuous sampling appears to be an acceptable approach for collecting random noise data.

Pulmonary impedance in dogs measured by forced random noise with a retrograde catheter.

Retrograde catheter and forced random noise techniques were combined to study the distribution of resistance and compliance in dogs following the inhalation of aerosols containing 2.5 and 5.0 mg/ml of histamine. Mean base-line peripheral resistance was 0.367 cmH2O . l-1 . s, agreeing with previous estimates. After correction for the endotracheal tube, the mean central airway resistance was 0.040 cmH2O . l-1 . s, considerably lower than previous estimates. This discrepancy was attributed to an overcorrection for the endotracheal tube resistance. The lower histamine dose caused a substantial increase in peripheral resistance, a relatively small increase in central resistance, and substantial decreases in total and peripheral compliance. After the higher histamine dose, changes in peripheral resistance and both compliances were similar to those obtained with the lower dose; however, the increase in central resistance was much larger than at the lower dose. The difference between total and peripheral compliance yielded estimates of airway compliance of 0.00306 l/cmH2O before and 0.00104 l/cmH2O after 2.5 mg/ml of histamine.

1726 TOTAL RESPIRATORY RESISTANCE IN CHILDREN WITH A HISTORY OF BRONCHIOLITIS

Children with a history of hospitalization for bronchiolitis (Br) as infants are reported to have decrements in lung function during later childhood; however, the effect of mild Br on lung function is not known. We measured the total respiratory resistance (Rrs) using the forced random noise technique in 32 children from a day care center who had their respiratory illness status monitored prospectively from birth. Nineteen children were observed to have had at least one episode of mild Br, while 13 had none. The mean height (Ht) ± SEM, mean Rrs ± SEM, and mean Rrs normalized for Ht (nl-Rrs) ± SEM are: The children with a history of Br have a significantly higher Rrs even when normalized for height than those without such a history. These data suggest that even mild Br early in life may be an antecedant of lung dysfunction later in life.

Random noise techniques in nuclear reactor systems

Random noise techniques in nuclear reactor systems

Soviet Physics—Acoustics

Reducing the vibration of smart structures based on adaptive active vibration control (AAVC) has been extensively studied in recent years, because it is a light weight and effective method for reducing low-frequency structural vibrations. The multi-channel AAVC methodology based on the filtered-X least mean square (FXLMS) algorithm is widely implemented in active control applications owing to its self-adjustment ability to adapt to dynamically varying structures. In this paper, a new multi-channel FXLMS with online secondary path modelling (SPM) is designed based on the auxiliary random noise technique. Through a defined indirect error signal, the proposed variable step-size (VSS) strategies can ensure that every online SPM filter and every active control filter have their own exclusive step size to update the coefficients. Moreover, the proposed auxiliary noise power scheduling (ANPS) strategy can ensure that the variation rules of auxiliary noise power applied to different secondary paths are different. A complete multi-channel AAVC real-time experimental system based on NI compact RIO is set up to conduct the experimental investigation. A series of AAVC control experiments on a piezoelectric smart cantilever plate with PZT sensors and actuators are conducted to demonstrate the validity and efficiency of the proposed method. The experimental results show that the vibration of the smart cantilever plate could be effectively attenuated with a high convergence rate. The proposed methodology has an important advantage in applications where active vibration control of piezoelectric smart structures is required.