**5. Cardiopulmonary signals at different sides: front, back, left, and right**

As the cardiopulmonary monitoring for patients and people under rubble requires heartbeat detection regardless their positions with respect to the system, measurements were performed at the four different sides from the PUT: front, back, left, and right. This section describes the measurement and shows the results of the contactless cardiopulmonary detection at four different sides.

#### **5.1. Measurement setup**

Performing a measurement begins by generating a continuous wave signal at the desired operational frequency. This CW signal, generated by the VNA, is driven to the transmitting antenna that is directed to the subject's chest. Reflected off the chest of the person under test, the signal is received by the receiving antenna and is driven back to the VNA. The phase variation *S*21, which corresponds to the difference in terms of phase between the transmitted and the received signal, is computed. The difference in phase is due to the chest displacement. Hence, it contains information about the cardiopulmonary signals when breathing normally

The theoretical values of the phase variation due to the chest displacement ranges and the average phase variation obtained by measurements when operating at 2.4, 5.8, 10, 16, and 60

For each of the utilized frequencies, the phase variation ranges within the theoretical limits. **Figure 2** shows the phase variations due to heartbeat signals detected at different frequencies and plotted within the same scale. It can be noticed that the phase variation increases when

**Figure 2.** Phase variation of *S*21 due to heartbeat signal measured when holding the breath at different operational fre‐ quencies: (a) 2.4 GHz, (b) 5.8 GHz, (c) 10 GHz, (d) 16 GHz, (e) 60 GHz, and (f) all frequencies over the same scaling.

**Frequency Wavelength (***λ***)** *∆θ* **for ∆***x* **= 0.2 mm** *∆θ* **for ∆***x* **= 0.5 mm Experimental** *∆θ*

2.4 GHz 125 mm 1.15° 2.88° 1.57° 5.8 GHz 51.72 mm 2.78° 6.96° 3.66° 10 GHz 30 mm 4.8° 12° 5.27° 16 GHz 18.75 mm 7.68° 19.2° 10.46° 60 GHz 5 mm 28.8° 72° 43.85°

and about the heartbeat signal when holding the breath.

42 Advanced Biosignal Processing and Diagnostic Methods

**Table 1.** Theoretical and measured phase variations due to chest displacement.

GHz are shown in **Table 1**.

The measurements were performed on a 54-year-old healthy subject, sitting at a distance of 1 m from the antennas. The operational frequency of the microwave system is 5.8 GHz with a total output power of 0 dBm. Each measurement lasts 30 s where the PUT breathes normally. The contactless measurement signal is acquired simultaneously with a PC-based ECG to be used as a reference signal for the heartbeat detection in order to validate the accuracy of both the microwave system and the signal processing technique. **Figure 3** presents the phase of *S*<sup>21</sup> measured at four positions: (a) measurement from the front side of the person, (b) measure‐ ment from the back side of the person, (c) measurement from the left side of the person, and (d) measurement from the right side of the person.

**Figure 3.** Phase variation of *S*21 due to the cardiopulmonary activities measurement at different sides from the subject: (a) front side, (b) back side, (c) left side, and (d) right side.

It can be noticed that the respiration signal is from the front side of the subject is clearer than other sides in time domain. The signal processing technique in the next subsection will show how it is possible to extract the heartbeat signal from the cardiopulmonary signal over all sides.

#### **5.2. Signal processing techniques**

Because the phase variations of *S*<sup>21</sup> caused by respiration are larger than those caused by the heart beating, processing techniques are required to extract heartbeat signal from the obtained cardiopulmonary signal. Previous works tend to apply the FFT in order to extract the heartbeat rate. This gives an average value of the HR over a specific window of time; hence, it lacks providing information about the variation of the HR in time and cannot be established in real time as it needs a long-duration window. In order to overcome these problems, the discrete wavelet transform (DWT) is applied to extract the heartbeat signal.

The DWT (*Wj* , *k*) of a signal *f(t)* is given by the scalar product of *f(t)* with the scaling function (i.e. the wavelet basis function *ϕ(t)* which is scaled and shifted:

$$(W\_j, k) \tag{4}$$

where the basis function is given by

$$\Phi\_{\parallel}, k(t) = 2^{\frac{-j}{2}} \phi(2^{-j}t - k) \tag{5}$$

where *j* is the *j*th decomposition level or step and *k* is the *k*th wavelet coefficient at the *j*th level [51]. DWT is computed by successive low-pass and high-pass filtering of the discrete timedomain signal [51]. DWT determination examines the signal at different frequency bands with different resolutions by decomposing the signal into approximation coefficients (A) and detailed information (D). Hence, this algorithm gives precise analysis of frequency domain at low frequency and time domain at high frequency. The DWT principle is resumed in **Figure 4**.

In general, *Dn* contains frequencies between *fs/2n* and *fs/2n+1*. As the HR varies between 60 and 120 beats per minute, the frequency of the heartbeat is located between 1 and 2 Hz. For the actual sampling frequency used in the VNA (666.7 Hz), no decomposition provides the signal having its frequency component between 1 and 2 Hz. Hence, a re-sampling is needed in order to convert the sampling frequency from 666.7 to 512 Hz. This lets the 1–2 Hz components be included in the 8th-level decomposition of the wavelet. Once the wavelet decomposition is extracted, the signal is reconstructed in time domain, and a peak detection method can be applied in order to detect peaks (beats), thus, to extract the heartbeat rate. The wavelet transform used in this study is the Bior2.4.

Position-Free Vital Sign Monitoring: Measurements and Processing http://dx.doi.org/10.5772/63915 45

**Figure 4.** Example of wavelet decomposition.

It can be noticed that the respiration signal is from the front side of the subject is clearer than other sides in time domain. The signal processing technique in the next subsection will show how it is possible to extract the heartbeat signal from the cardiopulmonary signal over all sides.

Because the phase variations of *S*<sup>21</sup> caused by respiration are larger than those caused by the heart beating, processing techniques are required to extract heartbeat signal from the obtained cardiopulmonary signal. Previous works tend to apply the FFT in order to extract the heartbeat rate. This gives an average value of the HR over a specific window of time; hence, it lacks providing information about the variation of the HR in time and cannot be established in real time as it needs a long-duration window. In order to overcome these problems, the discrete

> <sup>2</sup> , ( ) 2 (2 ) *j*

*<sup>j</sup> kt t k* -

*j*

where *j* is the *j*th decomposition level or step and *k* is the *k*th wavelet coefficient at the *j*th level [51]. DWT is computed by successive low-pass and high-pass filtering of the discrete timedomain signal [51]. DWT determination examines the signal at different frequency bands with different resolutions by decomposing the signal into approximation coefficients (A) and detailed information (D). Hence, this algorithm gives precise analysis of frequency domain at low frequency and time domain at high frequency. The DWT principle is resumed in **Figure 4**.

In general, *Dn* contains frequencies between *fs/2n* and *fs/2n+1*. As the HR varies between 60 and 120 beats per minute, the frequency of the heartbeat is located between 1 and 2 Hz. For the actual sampling frequency used in the VNA (666.7 Hz), no decomposition provides the signal having its frequency component between 1 and 2 Hz. Hence, a re-sampling is needed in order to convert the sampling frequency from 666.7 to 512 Hz. This lets the 1–2 Hz components be included in the 8th-level decomposition of the wavelet. Once the wavelet decomposition is extracted, the signal is reconstructed in time domain, and a peak detection method can be applied in order to detect peaks (beats), thus, to extract the heartbeat rate. The wavelet

, *k*) of a signal *f(t)* is given by the scalar product of *f(t)* with the scaling function

( ,) *W k <sup>j</sup>* (4)


wavelet transform (DWT) is applied to extract the heartbeat signal.

(i.e. the wavelet basis function *ϕ(t)* which is scaled and shifted:

**5.2. Signal processing techniques**

44 Advanced Biosignal Processing and Diagnostic Methods

where the basis function is given by

transform used in this study is the Bior2.4.

The DWT (*Wj*

**Figure 5** shows the result of the 8th-level decomposition upon applying Bior2.4 to the cardi‐ opulmonary signal detected from the front side of the PUT as well as the ECG signal extracted simultaneously with the cardiopulmonary signal.

**Figure 5.** ECG vs. cardiopulmonary signal extracted from the front side of the subject and processed using DWT Bi‐ or2.4 level 8.

Applying the peak detection method to both ECG and filtered cardiopulmonary signal detected at the front side from the subject gives, respectively, 41 R-waves and 44 peaks. This results in an HR of 85 bpm for the ECG and 93 bpm for the filtered cardiopulmonary signal.

The result of the 8th-level decomposition upon applying Bior2.4 to the cardiopulmonary signal detected from the back side of the PUT is shown in **Figure 6**.

**Figure 6.** ECG vs. cardiopulmonary signal extracted from the back side of the subject and processed using DWT Bi‐ or2.4 level 8.

Applying the peak detection method to both ECG and filtered cardiopulmonary signal detected at the back side from the subject gives, respectively, 42 R-waves and 42 peaks for both signals. This results in an HR of 88 bpm for the ECG and 87 bpm for the filtered cardiopul‐ monary signal.

The result of the 8th-level decomposition upon applying Bior2.4 to the cardiopulmonary signal detected from the left side of the PUT is shown in **Figure 7**.

**Figure 7.** ECG vs. cardiopulmonary signal extracted from the left side of the subject and processed using DWT Bior2.4 level 8.

Applying the peak detection method to both ECG and filtered cardiopulmonary signal detected at the left side from the subject gives, respectively, 40 R-waves and 38 peaks. This results in an HR of 82 bpm for the ECG and 80 bpm for the filtered cardiopulmonary signal.

The result of the 8th-level decomposition upon applying Bior2.4 to the cardiopulmonary signal detected from the right side of the PUT is shown in **Figure 8**.

**Figure 8.** ECG vs. cardiopulmonary signal extracted from the right side of the subject and processed using DWT Bi‐ or2.4 level 8.

Applying the peak detection method to both ECG and filtered cardiopulmonary signal detected at the right side from the subject gives, respectively, 42 R-waves and 38 peaks. This results in an HR of 87 bpm for the ECG and 80 bpm for the filtered cardiopulmonary signal.

#### **5.3. Results and discussion**

The result of the 8th-level decomposition upon applying Bior2.4 to the cardiopulmonary signal

**Figure 6.** ECG vs. cardiopulmonary signal extracted from the back side of the subject and processed using DWT Bi‐

Applying the peak detection method to both ECG and filtered cardiopulmonary signal detected at the back side from the subject gives, respectively, 42 R-waves and 42 peaks for both signals. This results in an HR of 88 bpm for the ECG and 87 bpm for the filtered cardiopul‐

The result of the 8th-level decomposition upon applying Bior2.4 to the cardiopulmonary signal

**Figure 7.** ECG vs. cardiopulmonary signal extracted from the left side of the subject and processed using DWT Bior2.4

detected from the back side of the PUT is shown in **Figure 6**.

46 Advanced Biosignal Processing and Diagnostic Methods

detected from the left side of the PUT is shown in **Figure 7**.

or2.4 level 8.

monary signal.

level 8.

Compared to the HR extracted from the ECG signal, the DWT with Bior2.4 family at decom‐ position level 8 shows accurate heartbeat detection. The signal processing technique applied to the signal extracted from the front side of the subject shows an error of 9%, while an error of 1% is obtained from the signal extracted at the back side of the subject. Also, the DWT applied to the signal extracted from the left side shows an error of 3% while an error of 7% is obtained for the signal extracted from the right side. **Table 2** shows the results in terms of HR calculation for both ECG and filtered cardiopulmonary signal for the four sides' measurements.

The heartbeat rate for both the ECG and VNA signals are calculated as follows:

$$HR = \frac{60\left(N - 1\right)}{d\_1 + d\_2 + \dots + d\_{N-1}}\tag{6}$$

where *N* is the number of peaks and *dk* is duration between two consecutive peaks. The peaks are detected using a classical peak-detection algorithm that is applied to both the ECG signal and the filtered cardiopulmonary signal.


**Table 2.** HR calculation for ECG and filtered cardiopulmonary signals and the obtained relative error.

The relative error of the HR is calculated as

$$Error = \frac{100^{\ast} \mid HR\_{ECG} - HR\_{INA} \mid}{HR\_{ECG}} \tag{7}$$

As shown in Section 5.1, the cardiopulmonary signal detected from the front side of the subject shows clear respiration signal in time-domain while other sides' signals do not. On the other hand, when applying the Bior2.4 family of the discrete wavelet transform to the cardiopul‐ monary signals, the highest accuracy in terms of heartbeat rate is obtained at the back side while the front side shows the lowest accuracy. This is due to the fact that chest displacement due to breathing is higher at the front side than other sides; hence, the chest displacement due to heart beating will be less affected by the respiration signal on other sides.

#### **6. Conclusion**

A microwave system used in order to detect the chest wall motion that contains information about respiration and heart beating is described. The system is tested at different operational frequencies: 2.4, 5.8, 10, 16, and 60 GHz on a subject at 1 m from the system while holding the breath for 10 s. Other measurements were performed at 5.8 GHz for different positions for the subject: front, back, left, and right sides. The first measurement is performed on a 27-year-old subject while holding the breath, while the second measurement is performed on a 54-yearold subject while breathing normally. Along with a PC-based ECG, measurements are performed with 0 dBm output power and for a duration of 30 s where the subject breathes normally. The proposed system shows the ability of detecting cardiopulmonary signals for the four sides' positioning: front, back, left, and right. Wavelet transformation is used in processing cardiopulmonary signals in order to extract the heartbeat signal. The 8th-level decomposition of Bior2.4 shows high performance in providing the heartbeat signal in time domain where high accuracy is obtained in terms of heartbeat rate. A peak detection method is applied to the reconstructed signal from the 8th-level Bior2.4 decomposition. Accuracy in terms of HR for different positions varies between 1% (from the back side) and 9% (from the front side). The proposed system and signal processing technique show the possibility of measuring the cardiopulmonary activities of the subject at four different positions with high accuracy. Compared to other studies, the proposed processing technique shows the ability of detecting the heartbeat signal in time domain, hence, giving information about the variation of the heartbeat rate in real time with an error less than 10%. Future work will concern performing measurements on persons with different ages and under different breathing circumstances as well as for persons in motion.
