Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Wearable BAMI device for acquiring three-channel PPG signals and three-axis acceleration signals.

(a) device, (b) example of measured PPG signal corrupted by low-intensity MAs, (c) example of measured PPG signal corrupted by high-intensity MAs, (d) power spectrum of signal (b) with true HR, and (e) power spectrum of signal (c) with true HR.

More »

Fig 1 Expand

Fig 2.

Time-frequency spectrum (TFS) of the PPG signals for subject 1 obtained by the BAMI device.

(a) true HRs (black circles) on the TFS, (b) dominant frequencies of the PPG signals (black circles) on the TFS, (c) dominant frequencies of the three-axis acceleration signals (black circles) on the TFS, and (d) dominant frequencies of the PPG signals (black circles) on the TFS after MA cancellation using the acceleration signals.

More »

Fig 2 Expand

Table 1.

Comparison of the HR estimation results obtained with and without MA cancellation for the ISPC (n = 23) and BAMI (n = 24) datasets.

The performance was evaluated on the basis of the mean absolute error (MAE).

More »

Table 1 Expand

Table 2.

Comparison summary of the HR estimation results obtained with and without MA cancellation.

Each result was obtained from all datasets (n = 47): ISPC (n = 23) and BAMI (n = 24) datasets. The performance was evaluated on the basis of the mean absolute error (MAE).

More »

Table 2 Expand

Fig 3.

Comparison of HR estimation results.

(a) with and (b) without the FSM framework for subject 1 in the BAMI dataset.

More »

Fig 3 Expand

Fig 4.

Sample HR estimation results obtained with MA cancellation.

(a) power spectrum PS(i) (blue solid line) and power spectrum PA(i) (yellow dotted line), (b) power spectrum PC(i) after MA cancellation, (c) Gaussian kernel-based modified power spectrum , and (d) power spectrum after MA cancellation based on .

More »

Fig 4 Expand

Fig 5.

Flowchart of the proposed algorithm.

More »

Fig 5 Expand

Fig 6.

Variation.

(a) MAE and (b) VHR with σ in the Gaussian function given by Eq (5).

More »

Fig 6 Expand

Fig 7.

Estimated HR trace comparison for two subjects of the ISPC dataset, obtained by DFDF (with MA cancellation only), FSM-DFDF (with MA cancellation and FSM framework), and FSM-SGPS (the proposed algorithm).

(a) DFDF results of subject 2, (b) FSM-DFDF results of subject 2, (c) FSM-SGPS results of subject 2, (d) DFDF results of subject 11, (e) FSM-DFDF results of subject 11, (f) FSM-SGPS results of subject 11.

More »

Fig 7 Expand

Table 3.

Performance comparison of the DFDF, FSM-DFDF, and FSM-SGPS HR estimation methods.

More »

Table 3 Expand

Table 4.

Comparison of MAEs of various HR estimation methods for the ISPC dataset.

More »

Table 4 Expand

Fig 8.

Bland–Altman plot and correlation.

(a) Bland–Altman plot of estimated and true HRs and (b) correlation for the proposed FSM-SGPS algorithm.

More »

Fig 8 Expand

Table 5.

Performance comparison of the DFDF, FSM-DFDF, and FSM-SGPS HR estimation methods for the 24 subjects of the BAMI dataset.

More »

Table 5 Expand

Fig 9.

Comparison of the estimated HR trace obtained.

(a) FSM-DFDF and (b) FSM-SGPS methods for one subject: (c) example of the measured 8-s PPG signal in region A, (d) reconstructed 8-s PPG signal after MA removal followed by inverse FFT, and (e) simultaneously measured 8-s ECG signal.

More »

Fig 9 Expand

Fig 10.

Examples of PPG signals corrupted by MAs.

(a)–(i) show PPG signals corrupted by different MAs. There are numerous MA patterns in PPG signals.

More »

Fig 10 Expand