eess.SPMar 20, 2026

Smartphone-Based Identification of Unknown Liquids via Active Vibration Sensing

Authors: Yongzhi Huang

Organizations: Information Hub, Data Science and Analytics (DSA) Thrust, The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Traditional liquid identification instruments are often unavailable to the general public. This paper shows the feasibility of identifying unknown liquids with commercial lightweight devices, such as a smartphone. The key insight is that different liquid molecules have different viscosity coefficients and therefore must overcome different energy barriers during relative motion. With this intuition in mind, we introduce a novel model that measures liquids' viscosity based on active vibration. However, building a robust system using built-in smartphone accelerometers is challenging. Practical issues include under-sampling, self-interference, and the impact of liquid-volume changes. Instead of machine learning, we tackle these issues through multiple signal processing stages to reconstruct the original signals and cancel out the interference. Our approach estimates liquid viscosity with a mean relative error of 2.9% and distinguishes 30 types of liquids with an average accuracy of 95.47%.

Figures & tables

Figure 1: (a) Using an iPhone 7 to detect a liquid. (b) The user interface of Vi-Liquid.
Figure 2: Physical measurement model of Vi-Liquid.
Figure 3: Feasibility-study setup with a separate motor, accelerometer, container, and alternating-current power supply.
Figure 4: Dissolving different masses of sucrose and NaCl in solution. (a) Change of viscosity. (b) Change of average peak value.
Figure 5: Distribution of different volumes of liquid in the mass–amplitude plane.
Figure 6: Measured β\beta and fτf_{\tau} , together with theoretical fτ′f^{\prime}_{\tau} .
Figure 7: The attenuation of volume to viscosity is not proportional.
Figure 8: Decaying-state signals of solutions with different concentrations after the motor stops.
Figure 9: Average peak value under different container materials and thicknesses.
Figure 10: The workflow of Vi-Liquid.
Figure 11: Transient signals at different sampling rates. (a) 1600 Hz. (b) 100 Hz.
Figure 12: Initial phone placement and direction.
Figure 13: Amplitude and SNR at different phone placements.
Figure 14: Undersampling problem and its solution. (a) Sampling the same phase repeatedly. (b) Phase-diverse sampling across periods.
Figure 15: The original signal and the magnitude spectrum obtained by sampling it at different time intervals.
Figure 16: Reconstructing the vibration signal with Supersampling Rate Reconstruction .
Figure 17: Amplitude–frequency curves of the vibration signal. (a) Original high-rate signal from BMI160. (b) Correctly re-ordered signal. (c) Incorrectly re-ordered signal.
Figure 18: Comparison between the reconstructed signal and the original high-rate signal.
Figure 19: Magnitude spectra of liquids with different volumes and different viscosities.
Figure 20: Using super-resolution reconstruction to align signals from different volumes.
Liquidβ\betafτ (N)f_{\tau}~(\mathrm{N})Vi-Liquid (cP)GT (cP)Error (%)
5% salt solution0.87750.01341.03 ±\pm 0.0191.003.00
10% salt solution0.87570.01421.09 ±\pm 0.0191.071.87
5% sucrose solution0.87600.01401.08 ±\pm 0.0251.061.89
10% sucrose solution0.87450.01551.19 ±\pm 0.0311.162.59
Coffee (10g sugar)1.37700.04983.83 ±\pm 0.1943.781.32
Coffee (15g sugar)1.43920.05254.04 ±\pm 0.0943.942.53
Table 1: Measuring viscosity for various liquids with Vi-Liquid and a viscometer ( ±\pm indicates standard deviation, and GT denotes ground truth).
Figure 21: Confusion matrix for liquid identification based on the estimated viscosity.
Figure 22: Estimated viscosity errors when straight-path interference is not removed.
Figure 23: Mean errors when the original under-sampled signals are used directly.
Figure 24: Viscosity variation before and after volume compensation.
Figure 25: Water contamination detection.
Figure 26: Concentration level estimation based on viscosity.
Figure 27: Exploring the upper boundary of viscosity measurement.
More in this topic

Classification

Classification maps observations to discrete labels, including familiar, unknown, hierarchical, or multiple categories.

See all 408 papers

Explore similar work

CardsList

  1. Structured-Token Attention for Fine-Grained mmWave Liquid Identification

    Oct 20, 2026Junseo Min, Joongheon KimClassificationAttention Mechanisms

    Department of Defense Convergence Technology, Korea University, Seoul, Republic of Korea · Department of Electrical & Computer Engineering, Korea University, Seoul, Republic of Korea

  2. Learning to Segment Liquids in Real-world Images

    Jan 2, 2026Jonas Li, Michelle Li, Luke Liu +2Image SegmentationBoundary-Aware Segmentation

    Department of Computer Science and Engineering, University of North Texas, Denton, TX 76203, USA

  3. Sound of Touch: Active Acoustic Tactile Sensing via String Vibrations

    Feb 18, 2026Xili Yi, Ying Xing, Zachary Manchester +1Tactile SensingRobotic Tactile Sensing

    University of Michigan · MIT