0.6 总9 VSkin 2cm 一Mus1c75d) Vithou defay selection -15 Speech (70dB) Without delay selection and EKF 1cm Music from the same speaker (65dB) 18 8 12 18 汤 12 16 Error(mm) Error(mm) Error(mm) (a)CDF for different algorithms (b)CDF for different diameters (c)CDF for different noise types 0.8 0. E25 ★查 20 Put on the table -Hold in hand 白白自 2 10 10 15 20 25 Error(mm) Speed(cm/s) Jamming distance(cm) (d)CDF for different use cases (e)Error for different speeds (f)Error for different jamming distances Figure 12:Micro benchmark results for movements hand only increases the average distance error by 3.42 mm, on average);iii)playing music from the same speaker that as shown in Figure 12(d). used by VSkin (65 dB on average).As shown in Figure 12(c) VSkin can reliably measure the movement distance with the average movement distance errors are 4.64 mm,5.93 mm speeds from 2 cm/s to 12 cm/s.We asked the user to move and 8.08 mm,respectively.Note that VSkin does not block his finger at different speeds for a distance of 6 cm.Figure the playback functions of the speaker. 12(e)shows the distribution of the movement distance errors with respect to the movement speeds.The average measure- 8.3 Evaluations on Touch Measurements ment error decreases from 11.00 mm to 4.64 mm when using VSkin achieves a touch detection rate of99.64%for different upsampling.Especially,when the moving speed is higher positions at the back of the mobile phone.We asked users than 8 cm/s,the average distance error decreases by about half,from 17.57 mm to 8.29 mm,when applying upsampling. to touch the back of the mobile phone for 100 times at the 11 different positions in Figure 11.VSkin missed only four This shows our upsampling scheme significantly improves touches among the 1100 tests.This gives a false negative the accuracy and robustness when the object is moving at rate of merely 0.36%.Since touching on the position close to high speeds. the speaker causes more significant changes in the structure- VSkin is robust to interfering movements that are 5 cm away borne signal,these four false detections are all at the position from the phone.To evaluate the anti-jamming capacity of 10 and 11 in Figure 11.VSkin also has low false positive ra- VSkin,we asked other people to perform jamming move- tios.When placed in a silent environment,VSkin made no ments,i.e.,pushing and pulling hand repeatedly at different distances,while the user is performing the movement.As false detection of touching for 10 minutes.When perform- shown in Figure 12(f),VSkin achieves an average move- ing jamming movements 5 cm away from the device,VSkin only made three false detections of touching for 10 minutes ment distance error of 9.19 mm and 3.98 mm under jamming Note that VSkin detects exactly the contact event,as users movements that are 5 cm and 25 cm away from the device, only moved their fingertip for a negligible distance in the respectively.Jamming movements introduce only a small touching experiments(measured air path length change of increase in the measurement error,due to the nice auto- only 0.3 mm).In comparison,touch detection that only uses correlation property of the ZC sequence that can reliably the magnitude of path coefficients has a lower detection rate separate activities at different distances. of 81.27%as discussed in Section 7.1. VSkin is robust to background audible acoustic noises and achieves an average movement distance error of6.22 mm under VSkin achieves an average accuracy of87.82%for classifying touches to three different regions of the phone.We divide the noise interferences.We conducted our experiments in three different environments with audible acoustic noises:i)an 11 different positions into three different classes as shown by different colors in Figure 11.We asked users to touch the indoor environment with pop music being played(75 dB on average);ii)a room with people talking being played(70 dB back of the phone at these 11 different positions for 100 times in each position.VSkin uses the delay of the structure path0 4 8 12 16 Error (mm) 0 0.2 0.4 0.6 0.8 1 CDF VSkin Without delay selection Without delay selection and EKF (a) CDF for different algorithms 0 4 8 12 16 20 Error (mm) 0 0.2 0.4 0.6 0.8 1 CDF 2cm 1.5cm 1cm (b) CDF for different diameters 0 4 8 12 16 20 Error (mm) 0 0.2 0.4 0.6 0.8 1 CDF Music (75dB) Speech (70dB) Music from the same speaker (65dB) (c) CDF for different noise types 0 4 8 12 16 20 Error (mm) 0 0.2 0.4 0.6 0.8 1 CDF Put on the table Hold in hand (d) CDF for different use cases 2 4 6 8 10 12 Speed (cm/s) 0 5 10 15 20 25 30 35 40 Error (mm) With upsampling Without upsampling (e) Error for different speeds 5 10 15 20 25 Jamming distance (cm) 0 5 10 15 20 25 Error (mm) (f) Error for different jamming distances Figure 12: Micro benchmark results for movements hand only increases the average distance error by 3.42 mm, as shown in Figure 12(d). VSkin can reliably measure the movement distance with speeds from 2 cm/s to 12 cm/s. We asked the user to move his finger at different speeds for a distance of 6 cm. Figure 12(e) shows the distribution of the movement distance errors with respect to the movement speeds. The average measure￾ment error decreases from 11.00 mm to 4.64 mm when using upsampling. Especially, when the moving speed is higher than 8 cm/s, the average distance error decreases by about half, from 17.57 mm to 8.29 mm, when applying upsampling. This shows our upsampling scheme significantly improves the accuracy and robustness when the object is moving at high speeds. VSkin is robust to interfering movements that are 5 cm away from the phone. To evaluate the anti-jamming capacity of VSkin, we asked other people to perform jamming move￾ments, i.e., pushing and pulling hand repeatedly at different distances, while the user is performing the movement. As shown in Figure 12(f), VSkin achieves an average move￾ment distance error of 9.19 mm and 3.98 mm under jamming movements that are 5 cm and 25 cm away from the device, respectively. Jamming movements introduce only a small increase in the measurement error, due to the nice auto￾correlation property of the ZC sequence that can reliably separate activities at different distances. VSkin is robust to background audible acoustic noises and achieves an average movement distance error of 6.22mm under noise interferences. We conducted our experiments in three different environments with audible acoustic noises: i) an indoor environment with pop music being played (75 dB on average); ii) a room with people talking being played (70 dB on average); iii) playing music from the same speaker that used by VSkin (65 dB on average). As shown in Figure 12(c), the average movement distance errors are 4.64 mm, 5.93 mm and 8.08 mm, respectively. Note that VSkin does not block the playback functions of the speaker. 8.3 Evaluations on Touch Measurements VSkin achieves a touch detection rate of 99.64% for different positions at the back of the mobile phone. We asked users to touch the back of the mobile phone for 100 times at the 11 different positions in Figure 11. VSkin missed only four touches among the 1100 tests. This gives a false negative rate of merely 0.36%. Since touching on the position close to the speaker causes more significant changes in the structure￾borne signal, these four false detections are all at the position 10 and 11 in Figure 11. VSkin also has low false positive ra￾tios. When placed in a silent environment, VSkin made no false detection of touching for 10 minutes. When perform￾ing jamming movements 5 cm away from the device, VSkin only made three false detections of touching for 10 minutes. Note that VSkin detects exactly the contact event, as users only moved their fingertip for a negligible distance in the touching experiments (measured air path length change of only 0.3 mm). In comparison, touch detection that only uses the magnitude of path coefficients has a lower detection rate of 81.27% as discussed in Section 7.1. VSkin achieves an average accuracy of 87.82% for classifying touches to three different regions of the phone. We divide the 11 different positions into three different classes as shown by different colors in Figure 11. We asked users to touch the back of the phone at these 11 different positions for 100 times in each position. VSkin uses the delay of the structure path