Speech extraction from vibration signals based on deep learning.

Extracting speech information from vibration response signals is a typical system identification problem, and the traditional method is too sensitive to deviations such as model parameters, noise, boundary conditions, and position. A method was proposed to obtain speech signals by collecting vibration signals of vibroacoustic systems for deep learning training in the work. The vibroacoustic coupling finite element model was first established with the voice signal as the excitation source. The vibration acceleration signals of the vibration response point were used as the training set to extract its spectral characteristics. Training was performed by two types of networks: fully connected, and convolutional. And it is found that the Fully Connected network prediction model has faster Rate of convergence and better quality of extracted speech. The amplitude spectra of the output speech signals (network output) and the phase of the vibration signals were used to convert extracted speech signals back to the time domain during the test set. The simulation results showed that the positions of the vibration response points had little effect on the quality of speech recognition, and good speech extraction quality can be obtained. The noises of the speech signals posed a greater influence on the speech extraction quality than the noises of the vibration signals. Extracted speech quality was poor when both had large noises. This method was robust to the position deviation of vibration responses during training and testing. The smaller the structural flexibility, the better the speech extraction quality. The quality of speech extraction was reduced in a trained system as the mass of node increased in the test set, but with negligible differences. Changes in boundary conditions did not significantly affect extracted speech quality. The speech extraction model proposed in the work has good robustness to position deviations, quality deviations, and boundary conditions.

Acoustic Insulation Characteristics and Optimal Design of Membrane-Type Metamaterials Loaded with Asymmetric Mass Blocks.

Membrane-type acoustic metamaterials (MAMs) are the focus of the current research due to their lightweight, small size, and good low-frequency sound insulation performance. However, there exists difficulties for extensive application because of the narrow sound insulation band. In order to achieve broadband sound isolation under the premise of lightweight, a novel MAM with asymmetric rings is firstly proposed in this paper. The sound transmission loss (STL) of this MAM is calculated by an analytical method and is verified by the finite element model. The different properties of the membrane when it is loaded with one, two, or four mass blocks are analyzed. The comparison with the traditional MAM proves the superior performance of this novel MAM. Moreover, by discussing the influence of the eccentricity and distribution position of the masses on the results, the tunability of the sound insulation performance of this MAM is proven. Finally, the Isight platform is used to optimize the MAM to further improve the broadband sound insulation performance: the average STL of the MAM is improved by 15.7%, the bandwidth above 30 dB is improved by 11.5%, and the mass density is reduced by 30.01%.

Sound Insulation Performance of Composite Double Sandwich Panels with Periodic Arrays of Shunted Piezoelectric Patches.

The existing sandwich structure of the aircraft cabin demonstrates a good sound insulation effect in medium and high frequency bands, but poor in the low frequency band. Therefore, we propose an infinite new lightweight broadband noise control structure and study its sound transmission loss (STL). The structure is an orthogonally rib-stiffened honeycomb double sandwich structure with periodic arrays of shunted piezoelectric patches, and demonstrates lighter mass and better strength than the existing sandwich structure. The structure is equivalent according to Hoff's equal stiffness theory and the effective medium (EM) method. Using the virtual work principle for a periodic element, two infinite sets of coupled equations are obtained. They are solved by truncating them in a finite range until the solution converges. The correctness and validity of the model are verified by using simulation results and theoretical predictions. Eventually, a further study is performed on the factors influencing the STL. All the results demonstrate that the STL in low-frequency can be improved by the structure, and the sound insulation bandwidth is significantly broadened by adding shunted piezoelectric patches. The structure can provide a new idea for the design of broadband sound insulation.

A comparative investigation on absorption performances of three expanded graphite-based complex materials for toluene.

Three kinds of expanded graphite-based complex materials were prepared to absorb toluene by dispersing plant oil, animal oil and mineral oil on the surface of expanded graphite, respectively. These complex materials were characterized by scanning electronic micrograph, contact angle meter and Brunauer-Emmett-Teller surface area. And their absorption capacities for toluene were comparatively investigated. The results showed that the surfaces of the three types of sorbents were very hydrophobic and nonporous, but they all had excellent absorption capacities for toluene. And their absorption capacities were proportional to the toluene concentration in streams and decreased differently with increasing the absorption temperature. It was noteworthy that the absorption capacities varied with the unsaturated degree of the complex materials and kept unchanged under different relative humidities of streams. Moreover, the regeneration experiments showed that after 15-run regeneration the absorption capacities of expanded graphite modified by mineral oil almost kept unchanged, while that of expanded graphite loaded plant oil and animal oil dropped by 157 and 93.6 mg g(-1), respectively. The losses of their absorption capacities were ascribed to the destruction of their unsaturated carbon bounds.