Calculate phase difference between two general signals (e.g. HbO and Hb) using Hilbert transform

1 min read

In a recent fNIRS journal club (vedio recorded here), Dr. Tong talked about their work on the phase difference between oxy and deoxy Hb, and its relationship with participants’ age. This article is a demo of how to use Hilbert transform to calculate the phase difference between two signals, and whether it’s valid if the signals contain a wide range of frequencies.

Tool: We use MatLab’s hilbert function to calculate the instant phase of a signal. The code can be found in the end of this article.

Case 1: single frequency

Here we tried the simplest case: x and y are single frequency signals with phase difference pi/2, as demonstrated in the figure below. We expect that the calculated phase difference is close to pi/2 or 1.57.

As expected, the phase difference is pi/2! Hilbert transform is doing well in picking up the phase difference! It is noted, however, the calculation is more robust in the middle of the signal.

Case 2: 2 frequencies

What about more complex signals? If x and y both contains two frequencies, each with phase difference pi/2, will Hilbert transform find the correct value?

Yes. It turns out Hilbert transform works for two-frequency signals as well.

Case 3: a lot of frequencies

In this case we generated x and y each with 50 random frequencies. The resulted signals are essentially random:

Again, Hilbert transform found the correct phase difference, pi/2:

People may argue that in such a wide-band case the “phase difference” is not meaningful. However, if the phase difference for each frequency is similar, then it is reasonable to say that the phase difference between the overall signals exists and is that value.

Code (MatLab)

Update: in the original version I shift phase manually (and contains an error!). Now we use unwrap and wrapToPi, suggested by Yafeng Pan, which is easier to use and produce the correct phase shift. Thank Yafeng!

% Demo of Hilbert transform to calculate phase difference between two
% signals
% Xu Cui 2021/6/7

% single frequency
t = [0:0.01:100];
x = sin(t);
y = sin(t+pi/2);
figure; plot(t, x, t, y);
legend('x','y');

xh = hilbert(x);
yh = hilbert(y);

xphase = unwrap(angle(xh));
yphase = unwrap(angle(yh));

figure; plot(t, xphase, t, yphase,'.');

phase_diff = wrapToPi(xphase-yphase);
figure; plot(t, phase_diff, '.');

% 2 frequencies
t = [0:0.01:100];
x = sin(t) + sin(2*t);
y = sin(t+pi/2) + sin(2*t+pi/2);
figure; plot(t, x, t, y);
legend('x','y');

xh = hilbert(x);
yh = hilbert(y);

xphase = unwrap(angle(xh));
yphase = unwrap(angle(yh));

figure; plot(t, xphase, t, yphase,'.');

phase_diff = wrapToPi(xphase-yphase);
figure; plot(t, phase_diff, '.');

% a lot of frequencies
t = [0:0.01:100];
x = zeros(size(t));
y = zeros(size(t));
frequency = rand(1,50)*10;
for ii=1:length(frequency)
    f = frequency(ii);
    x = x + sin(f*t);
    y = y + sin(f*t+pi/2);
end

figure; plot(t, x, t, y);
legend('x','y');

xh = hilbert(x);
yh = hilbert(y);

xphase = unwrap(angle(xh));
yphase = unwrap(angle(yh));

figure; plot(t, xphase, t, yphase,'.');

phase_diff = wrapToPi(xphase-yphase);
figure; plot(t, phase_diff, '.');



写作助手,把中式英语变成专业英文


Want to receive new post notification? 有新文章通知我

第四十五期fNIRS Journal Club通知2023/9/23, 10am 张宗博士,侯鑫博士

近红外脑功能成像(fNIRS)是脑科学与脑疾病研究的新兴技术手段。对于fNIRS脑成像数据中空间与时间信息的可视化、解读与统计分析是fNIRS研究的重要环节。北京师范大学朱朝喆教授团队在十余年研究成果
Xu Cui
9 sec read

第四十四期fNIRS Journal Club视频 牛海晶教授

Youtube: https://youtu.be/gOQVFQ-LH7w Youku: https://v.youku.com/v_show/id_XNjAxMDg2MDYwNA==.html 牛海
Xu Cui
5 sec read

第四十四期fNIRS Journal Club通知2023/8/26, 10am 牛海晶教授

在儿童和早期青春期期间,大规模功能网络中的脑信号复杂度(BSC)是如何涌现和演变的,以及是什么因素塑造了这些动态?来自北京师范大学的牛海晶教授将分享他们利用静息态近红外技术研究儿童发育的相关工作。 时
Xu Cui
6 sec read

One Reply to “Calculate phase difference between two general signals (e.g. HbO…”

Leave a Reply

Your email address will not be published. Required fields are marked *