More and more people turn to online voice communication with loudspeaker-equipped devices due to its convenience. To prevent speech leakage, soundproof rooms are often adopted. This paper presents mmPhone, a novel acoustic eavesdropping system that recovers loudspeaker speech protected by soundproof environments. The key idea is that properties of piezoelectric films in mmWave band can change with sound pressure due to the piezoelectric effect. If the property changes are acquired by an adversary (i.e., characterizing the piezoelectric effect with mmWaves), speech leakage can happen. More importantly, the piezoelectric film can work without a power supply. Base on this, we proposed a methodology using mmWaves to sense the film and decoding the speech from mmWaves, which turns the film into a passive "microphone". To recover intelligible speech, we further develop an enhancement scheme based on a denoising neural network, multi-channel augmentation, and speech synthesis, to compensate for the propagation and penetration loss of mmWaves. We perform extensive experiments to evaluate mmPhone and conduct digit recognition with over 93% accuracy. The results indicate mmPhone can recover high-quality and intelligible speech from a distance over 5m and is resilient to incident angles of sound waves (within 55 degrees) and different types of loudspeakers.

In a densely distributed mobile crowdsourcing system, data collected by neighboring participants often exhibit strong spatial correlations. By exploiting this property, one may employ a portion of the users as active participants and set the other users as idling ones without compromising the quality of sensing or the connectivity of the network. In this work, two participant selection questions are considered: (a) how to recruit an optimal number of users as active participants to guarantee that the overall sensing data integrity is kept above a preset threshold; and (b) how to recruit an optimal number of participants with some inaccurate data so that the fairness of selection and resource conservation can be achieved while maintaining sufficient sensing data integrity. For question (a), we propose a novel task-centric approach to explicitly exploit data correlation among participants. This subset selection problem is regarded as a constrained optimization problem and we propose an efficient polynomial time algorithm to solve it. For question (b), we formulate this set partitioning problem as a constrained min-max optimization problem. A solution using an improved version of the polynomial time algorithm is proposed based on (a). We validate these algorithms using a publicly available Intel-Berkeley lab sensing dataset and satisfactory performance is achieved.

For comparison of noise levels between superconducting gravimeters (SGs) and gPhone gravimeters, we conducted an experiment in which we carried out co-located observations of gravity changes by the same gPhone gravimeter and Superconducting Gravimeter (SG) at Wuhan, Lhasa and Lijiang stations respectively, and the observation at each station lasted approximately 2 months. We calculated the root mean square (RMS) of the gravity residual in each station and selected the five days in which the gravity residuals had the minimum RMS representing the background noise level as a quiet period. Based on the gravity residuals in these five quiet days, we calculated the mean power spectral density (PSD), seismic band noise magnitude (SNM) and sub-seismic band noise magnitude (SSNM). Furthermore, we simulated the influence of scale-factor error on SNM. We found that the SNM of the SGs ranged between 1.22 and 1.50 and the SSNM ranged from 3.96 to 4.74 during the experiment. The SNM of the gPhone gravimeter ranged from 2.93 to 3.06, and the SSNM of the gPhone gravimeter ranged between 4.17 and 5.29. The SNM of the SG at Wuhan station was the highest among three SGs, while the corresponding SSNM was the lowest. The SG at Lhasa station, in contrast, had the highest SSNM and lowest SNM. The PSDs of gPhone gravimeters in high-frequency bands, including the seismic bands, were much larger than those of SGs. In the sub-seismic bands, SGs can approach or reach seismograph detection level. In the band frequency below $3\times {10}^{-5}\text{Hz}$, SGs and gPhone gravimeter have lower PSDs than seismometers, indicating that high-precision continuous gravimeters are better than seismometers in detecting long-period signals. The variation of SNM caused by the scale-factor is generally in the range from −0.092 to 0.083, which has little influence on SNM result.

Inter-seasonal and geodynamics-related gravity changes are important geoscientific signals that are extractable from gravimeter observations after removing background information as local hydrology gravity effect. With two superconducting gravimeters (SGs: OSG-053 and iGrav-007) located in different tectonic units, continuous global navigation satellite system data and absolute gravity observations, Wuhan, China, is an ideal location for investigating the effects of gravity resulting from significant local hydrology mass variations. We have processed approximately 26 months of gravity data collected from the SGs in Wuhan and obtained residuals of −40 nm·s−2 for OSG-053 and 100 nm·s−2 for iGrav-007. The hydrological observations indicate an estimated gravity increase of 42 nm·s−2 near iGrav-007, which mainly results from an increase in unconfined water level with an aquifer-specific yield of approximately 0.1. However, the gravity changes around OSG-053 are mainly from soil moisture and reach −90 nm·s−2. The soil type, thickness, and water content parameters are obtained from hydrogeological surveys and drilling data. The deep confined water level rises by 2.5 m, which introduces a 1 nm·s−2 gravity variation with a specific storage approximately 0.00001 from the field unsteady-flow pumping test. The modeled gravity is approximately −40 nm·s−2 around OSG-053 and 90 nm·s−2 around iGrav-007, in accordance with the observed gravity variations. The difference in gravity changes between the two SG observations can be explained by different local water storage environments. Our results suggest that unconfined and soil water significantly impact the in-situ gravimetry, and that further detailed hydrogeological surveys are required. A combined investigation of gravity and water levels can be a useful approach for monitoring aquifer storage conditions and groundwater management.

The North–South Seismic Belt was analyzed using gravity observation data from 2011 to 2015, and the nontidal analysis results show that there was a nonlinear gravity change at both the Chengdu and Guza seismostations one month before the Leshan M5.0 earthquake