Signal Analysis using Matlab - A Heart Rate example

[@blakemartin4820]

This is a great tutorial. I really like the way you show how errors happen and how to fix them. I also like your very clear coding language. My one suggestion is that MatLab is quite slow in doing loops, so it is better to use logical indexing. The code remains almost identical, but rather than an 'if' loop, I just make a variable called 'beat_grab' that grabs all of the counted beats at one time: %% count dominant peaks beat_grab = sig(k)>sig(k-1) & sig(k)>sig(k+1) & sig(k) >1; beat_count = length(beat_grab); Then continue with code to divide the beats Original code runs in about 0.001754 seconds. Altered code runs in about 0.000208 seconds, ten times faster. This may not matter depending on your sampling rate and the length of your data, but MatLab programmers generally try to be as efficient as possible, and loops are low on effeciency. Still I learned some new things too! Thanks.

[@ddorran]

I tend to avoid loops myself but I use this tutorials to teach programming as well and not all programming languages work in the same way as matlab - but they do all have loops! Thanks for your comments!

[@TheKarantheking]

Thank You very much. This video is still very helpful in 2022 !!

[@ddorran]

Here's hoping it'll still be useful in 2032!

[@lparry92]

Thank you so much for providing an example with data that can be attained too. It helps so much to be able to follow this tutorial with you

[@ddorran]

No problem! Glad it helped.

[@ketakipatil7536]

Dear David Sir, Found it very helpful. Thank you so much

[@ddorran]

Nothing on general feature extraction yet but "Matlab Signal Analysis - frame by frame analysis of a signal" might be of help if you are starting out. Also there's one called "Frequency Domain 1.3 - Demonstration - Application of frequency domain view " that shows how to extract frequency domain info that might be useful.

[@ddorran]

you can download it from the web. Can't provide url's here but check out the firs 30 seconds or so of the video

[@anuj.zest-AnujKhandelwal]

Thanks David, "Frequency Domain 1.3 - Demonstration - Application of frequency domain view" was really helpful to understand the concept. Yes, I have frequency-domain and time-domain signal data and I have to perform feature extraction. I am still trying to figure out the code apply to my dataset.

[@Ikopif1]

Hi David. The code looks good. This will be very helful for my analyses. I appreciate your fast respone. Thank you very much!

[@mistbabe26]

Your amazing! Thank you, Continue to do the great work. You helped me a lot for my term project.

[@ClaudinneV]

Thanks, very useful. I'm new in Matlab and this video is helping me a lot, I need to create a tachogram based on the ECG data,, I think I can do that, since with your code the prominant peak values have been obtained

[@khurin3938]

sir , why double layer winding is used in electrical machine instead of single layer winding?

[@coddy57]

Perfect! It works fine. That's what I'm talking bout. Thanks a lot you Mr. Dorran :)

[@oliviamccormick2064]

Hi, great video but is the findpeaks function a good way to be able to find the peaks of the P and T wave? Obviously, I need to exclude the R wave but I have been playing around with the findpeaks command and just can't seem to figure it out. Any suggestions?

[@african_]

Hi Dorran, I am doing research on using remote photoplethysmography to detect 3D masks and prevent presentation attacks. I am wondering whether you have any tutorials regarding that topic. I am honestly struggling to find such tutorials showing the implementation of such models. There's one specific method for extracting signals from facial videos called CHROME but I can't find any video showing how such code is written and implemented so that I can find ways to improve after understanding it. Would you be kind to help me, please? Your videos show that you write such code flawlessly and it's easy to understand. Please help, Dorran. If you have a source you can point me to, I'll be very grateful. Thank you so much.

[@ddorran]

That didn't format as intended. You'll probably have to put a newline after "while 1"

[@sofidesima1359]

Hi Davis, I need to know the cutoff frequency and the sampling frequency of the ECG signal, because I need to make a filter. thanks

[@mustafahashim1906]

it's really was very useful tutorial

[@myonlineteaching3755]

hi david.. glad to find your channel.. can u do a tutorial on various features that can be extracted in signal processing such as mfcc etc. thnks!

[@ddorran]

I'll put them on the "to do" list mimi - won't be any time soon though. The list is quite long.

[@myonlineteaching3755]

David Dorran very nice of sure.. sure.. illw wait. :)

[@albertguimfack6360]

Thanks you very much for this video. Very attractive.

[@LVK520]

Hi Mr, I would like to calculate the heart rate of the audio sound get from a microphone with a sample rate of 48000, is that code can work with? need your help and thank you.

[@ddorran]

Its easy enough. Here's some code that plots the signal from a sound card in real time (with thanks to my colleague Richard Hayes!). Fs=11025; N=2048;t=(0:N-1)/Fs; while 1 y = wavrecord(N,Fs,2); plot(y); axis( [0 N -1 1]); drawnow; end

[@kwesi4812]

Please How do i use the loop function to get the change in Torque signal at angles 5°, 10° ,15° & 20° from an ankel dynamometer

[@coddy57]

Hi,I designed a circuit with a stethoscope, the output of my circuit goes to the sound card of my PC, the matter is: is it possible to get that output signal (heartbeats) from stethoscope in real time using MATLAB? Is it difficult? I really appreciate your answer. Thank you!

[@rrahimi918]

how can I store the picks in a function in order to get R to R distance of ecg signal?.Its possible to calculate the HR from R-R diffrence.thank you very much

[@hidourisami578]

thank you ..! this tutorial has been very helpful.. but i work with polar team, so I only focus on heart rate variability (the RR intervals). Can you help me to have a code to process the HRV signal recorded with a polar team?

[@khashablanca]

Great video! i have a query that i was hoping you would be able to help me with, for my final year project i have been researching into calculating distance using sound on an iphone 6. I have been playing short frequency sweeps on one iphone and recording the data on another phone sitting on top of the other. What i'm planning to do is calculate the delay between the initial sound and the reflected sound and combining that with the speed of sound to give me the distance between a wall and the iphone. however i'm struggling to do so. I know you are a wizard on MatLab and was wondering if there was any techniques or methods to approach in calculating that time delay within MatLab or Audacity.

[@dindardosky6434]

Thank you very for this valuable video, really I good much information from it.

[@fuhbigbiufb]

I got a signal .mat and I plotted that on matlab. I know to show de number of samples , but i don't know the sampling frequenci . did you know if is possible to know this with only de doc of de signal?

[@harfenafaren9679]

hi, can you tell me the name of this ECG ecg sample you took from physionet?

[@ocayaro]

Great video. However it is "prominent", not "prominant". Also, ideally the average heart rate should be a running average, taken from the first two prominent peaks. That way two values are given, an instantaneous BPM and a running average BPM. For a critical care patient 1-minute makes the difference between life and death.

[@manyeaphooi5768]

Dear David, Really is a nice video and I learn a lot from here. Thanks a lot. But I'm just curious can be the frequency sampling rate be changed? Or it must be fixed at 100Hz? Best Regards, Hooi Man Yeap

[@ddorran]

The ecg data was originally captured at this rate (100Hz). It could have been captured at a different rate.

[@manyeaphooi5768]

David Dorran Okay, get it... Thanks for your information. :D

[@cooldudeniks1]

Hey do u have PPG waveform as u had the ECG waveform and its values in txt format.?? I am working on the same thing but using PPG waveform..can u help me getting it..thank you

[@binujanardhanan1299]

Nikhil Goregaonkar www.physionet.org/cgi-bin/atm/ATM

[@anilpandey3579]

I liked this video. What if we just want the prominent peaks to be plotted/marked. What kind of modification in the code would be necessary in such a condition?

[@ddorran]

The following should work by highlighting prominent peaks with red circles. beat_count = 0; peak_indicator = ones(1, length(sig))*NaN; for k = 2 : length(sig)-1 if(sig(k) > sig(k-1) & sig(k) > sig(k+1) & sig(k) > 1) %k peak_indicator(k) = sig(k); %disp('Prominant peak found'); beat_count = beat_count + 1; end end plot(sig); hold on plot(peak_indicator,'ro');

[@chikaokolo4929]

Excellent teacher

[@luizgabriel4110]

lindão salvou o dia :)

[@kellysheerin3530]

Hi David, Thanks a lot for posting this, it has been very useful for someone new to Matlab. I've adapted the code you've presented to help analyse my acceleration data. However, what I'm really after are the magnitudes of the peaks, which in your example would correspond to the magnitudes of your 'R' wave. Do you have anything that would be helpful for me in this regard? Thanks!

[@ddorran]

Do you just want to display the magnitude values associated with each peak? If so its just a case of replacing "beat_count = beat_count + 1" with "abs(sig(k))". Not sure if I understood the question fully.

[@kellysheerin3530]

David Dorran That was exactly what I was after... Thanks!

[@deit1813]

Hi, if l want to store the values of k in an array how should i write? thanks

[@alisahouskova]

I'd like to know, how to create noise, which can remind for example: pacient's moving, not correctly plugged electrodes etc. ... any ideas? Thx a lot.

[@antoniodejesusramirezcruz4040]

you helped me a lot thnk u

[@realcountry2054]

2:02 is it time-variable data or spatial data?

[@evandovich]

Hi Guys, Is it possible to to plot the prominent peaks as graph. My system keeps running out of space when i try. Any solution of code given will be much appreciated. Cheers

[@maxmendez6173]

Hi man that was really heplful!! how can I set a threshold for the peaks?

[@Judiciary-100]

sir can you show how empirical mode decomposition work using matlab code

[@Tobchi2010]

Very nice video. I give you an up thumb

[@ashidilkhan]

very very nice explanation

[@solomoon100]

Hello David do you have video on system identification?

[@jesalshah1309]

thank you ..! this tutorial has been very helpful..!! though i wanted to know if you can segregate the qrs complex from any of the ecg waves using matlab..?? m and yes any ecg wave means it can also be from the subject having heart rate disorders.

[@migly0105]

How can I store the i values that contain peaks?

[@afandydwiahmad5482]

hello sir i interest about your explanation. can i detect pqrst signal

[@danishbebal6708]

I m not getting the value of k...will u plzz help me

[@FuckenMoonFace]

Sir @David Dorran this signal is noise free isn't it?

[@gehadsamir5663]

how exactly did you know the frequency of the sample??? (100 hz)?? I've searched for a long time and can't seem to find where exactly is that information about the sample

[@ddorran]

That was provided with the data when I got it.

[@gehadsamir5663]

@@ddorran thank youuu

[@ronitnath3786]

Respected Sir...How to obtain the sample values from physionet.org. I have searched the physionet.org database and the signals are stored in .dat form. Could u please help me out in retrieving the sampling points in the signals?? I need it urgently for my final year project.

[@ddorran]

I have a video on reading an ECG signal (search reading ECG signal on my channel - links not allowed I'm afraid). From 2:30 to 6:00 is most relevant to this issue.

[@asifraj321]

Very useful and important tutorial. How can I reconstruct the signal for 100 bpm by using the Fourier series for the same volume flow rate?. Hope you reply soon. Thank you

[@ddorran]

I'm afraid I don't understand exactly what you mean. Can you clarify?

[@asifraj321]

@@ddorran Actually I have a velocity waveform corresponding to 75 bpm and with reference to it, I want to construct another waveform for a different heart beat like 100 bpm. Hope it clears now

[@ddorran]

@@asifraj321 Could you just resample it?

[@joshuamichael6371]

Thanks David. David can you tell me how to detect the QRS complex from the ECG wave? i.e. how do i extract the QRS from the PQRST Waveform?

[@ddorran]

I'm not sure how it's done in practice. I'm sure there are numerous publications on how you might proceed. My own thoughts would be to identify the zero crossings each side of the dominant peak and then just extract fixed amount each side of the zero crossing as a percentage of the number of samples between those zero crossings. This would seem easy to code up but may not be suitable for the application you have in mind.

[@dragonoidpop5]

Hi Joshua, did you found the answer to your question? Actually im doing a proyect and I need that too, Ill appreaciate very much your help

[@dersstri]

Nice video. Any references you used for making this video would be appreciated

[@shuangzhao5575]

Very helpful! Thanks a lot!

[@parijatprakash]

Can this same code work on an EEG signal too ?

[@19va83]

Thanks a million!! that was wonderful!

[@krishnaneupane9354]

sir please provide how to plot the HRV signals and how to bandpass the ECG signal using 50Hz bandwidth and also for MIT-BIH arrythmia database

[@hayatmoussaoui9923]

Hi....I've a Matlab script about peaks detection , the problem that I can't understand how it works :( . can i show it to you ?

[@chshaban4696]

@moussaoui hayat yes u can show

[@Jaked_chan]

gracias pana

[@Ikopif1]

Hi David. I would like to know how to store all instant of times whre the prominent peaks happened. Could you explain me how to do that. Thanks for your video.

[@ddorran]

I've modified some code I posted before to include a new vector called peak_times which is updated every time a prominent peak is found. This should do the job. fs = 100; beat_count = 0; peak_indicator = ones(1, length(sig))*NaN; for k = 2 : length(sig)-1 if(sig(k) > sig(k-1) & sig(k) > sig(k+1) & sig(k) > 1) %k peak_indicator(k) = sig(k); %disp('Prominant peak found'); beat_count = beat_count + 1; peak_times(beat_count) = k/fs; end end plot(sig); hold on plot(peak_indicator,'ro');

[@arisunny2489]

Hi i have questions and i have project about this How can i be in touch with you?

[@tommy11492]

hi david! i am working on datas that i' ve taken from physionet.org, that are ina a .dat extension. i can't load them on matlab, because they contain also notASCII-symbols. what can i do??

[@ddorran]

Check out physionet.org/tutorials/physiobank-text.shtml. This should be of help.

[@ipeknurlu3849]

which data did you use as raw ecg data from physio.net?

[@ketakipatil7536]

dear david sir, Can I get help in calculating PR interval?

[@josephEgypt]

Thanks a lot for this video. I understood very well. Could you help me with ECG signal noise filtering using matlab.

[@aramaeslam4966]

Hello, did you get access to the signal filter?

[@YahYaAlabrash98]

Amazing!! is there a method so that we can detect the Respiration Rate form the ECG Signal?

[@sirranhaal3099]

ECGs carry no information about the respiratory rate.

[@sayajujur2565]

Dear David, I am wali, i am a PhD student in Malaysia. Thank you for this video. It is great to start ECG signal processing. I wonder , if you could share the codes in txt format ?! I want to modify it to R peak detection algorithm. My signals is corrupted. so before R peak detection, i want to add at least 3 filters. This will be based on your code. Kindly, Wali

[@ddorran]

I'm afraid I don't have that code any more but you should be able to get it from the video.

[@sayajujur2565]

thank you sir

[@ibrahimfatnassi]

@@sayajujur2565 Can you please send me your email id

Good one. Thank you!

[@toncanan]

I was wondering as to how you got the integer values for your ecg.txt

[@ddorran]

I downloaded the signal from physionet.org

[@jawadghafoor5]

You can use different sensors to get digital values, like 'Ad8232' is a very good ECG sensor.

[@brashid421]

Hey,i would appreciate if u help me in making my project.i want to make a device that senses our heart beat .first it calculates the normal heart beat when it comes in touch with human body(even if the person wore a thick layer of clothes underneath) then when the heart rate is decreasing it will give an alarm.if the hear rate increase within 2 secs then the alarm will stop,if not then the volume of alarm increases. If u make a video on it ,plz do let me know about its -Components -Expenses -And if any coding needed If u wanna know about it in a more detailed way,then plz email me

[@sajpoe1235]

i didnt understand what BPM is

[@anuj.zest-AnujKhandelwal]

any help with "FEATURE EXTRACTION" ?

[@SudathPiyasena]

thank you very much

[@hanenbenjomaa2544]

can u help me getting it format.txt

[@ZainAbbas12]

I really like your videos, I am working on feature extraction of EEG signals...If possible can you do some signal processing such as Analog to digital conversion of EEG signals or some form EEG feature extraction using wavelet transform or autoregressive moving average method or some other. Actually i have difficulty modelling the parameters.

[@ddorran]

For analog to digital conversion all I have are videos on aliasing and quantisation issues. I don't have any expertise with EEG signal modelling but intend to do some videos of ARMA based modelling for speech when I get a chance. Best of luck.

[@ZainAbbas12]

Okay thankyou david. I will check on aliasing and quantisation issues. And looking forward to ARMA example video

[@enngennng5633]

Hi! It is very good, I will give you 5 stars. Do you have data for PPG signal if so please send.

[@MrArammos]

How can i convert and .dat file to run it on Matlab??

[@ddorran]

.dat files are not standardised and could relate to lots of different formats. See if you can open the file with a text editor to see how the data is structured. This might not be possible though. If the data is stored as a sequence of numerical values you load in the data using the load command. data = load('filename.dat');

[@wdyangcanjie123]

thanks for sharing

[@hafizatulamirah5866]

do you know where i can get database .mat file for normal ecg signal?

[@ramakrishnadesala7631]

physionet.org

[@ramakrishnadesala7631]

physionet.org

[@manzweichipungu5838]

Can you please specify the database because there is more of abnormal signals than normal.

[@krishnaneupane9354]

can you also provide me a some documents of your project, i'm also doing project on physionet MIT-BIH arrhythmia database

[@gayansanjeewa9123]

good job

[@toncanan]

are all or most peaks of the heart beat always greater than 1...?

[@ddorran]

only in this example. The amplitude values are dependent upon how the signal was initially captured and can vary significantly. Also, in practice there are other issues such as DC drift that can occur and do cause additional problems.

[@toncanan]

David Dorran is there a possibe formula to determine that? since during the detection of dominant peaks, you compared 1 with every signal..

[@ddorran]

Anthonette Cantara you could try finding the maximum of the signal and compare against 0.8*max_val, for example. This wouldn't work if there was DC drift though but you could filter this effect out using a high pass filter or maybe just update the max_val every few cycles.

[@toncanan]

is this method of calculation using rr interval?

[@toncanan]

+David Dorran

[@gukesu3864]

i did like you,but i got beat_count = 10 BPM = 120

[@alisahouskova]

Hey man, fuckin´ great! I love your channel. My bachelor thesis is about signal analysis (ICA over ECG).

[@ddorran]

noise = randn(1, 10000); %10000 samples of broadband random noise fs = 500; Fnyq = fs/2 [b a] = butter(5, [45/Fnyq 55/Fnyq], 'bandpass'); noise_band_50Hz = filter(b,a, noise);

[@mchandragaddam]

very needy !!

[@gayansanjeewa9123]

this verry nice

[@otsilekgaladua5485]

thank you so much sir. one question: is the fs always = 100Hz for this particular signal regardless of the length(duration of the signal taken from this site www.physionet.org/cgi-bin/atm/ATM ?

[@ddorran]

+kutlo kgaladua ECG signals can be sampled at different rates. 100Hz and 128Hz appear to very common but higher sampling rates are also used. The sampling rate is is constant for each signal - it doesn't matter what the duration is the sampling rate remains the same.

[@venugopal-gi9jt]

also any imported files in code please

[@ddorran]

+venu gopal I'm afraid I don't have the code saved anywhere. A link to the imported ecg signal is shown at the start of the video

[@dwadeLBJ2336]

York U 1011 students please stand up

[@SuperMowZ]

"findpeaks" is a bit easier :-P ....

[@marcinwojciechowski4938]

(y)

[@JuanAntonio-ff4mg]

porque diablos no hablan en español!!!!!.

[@robosumo5670]

Níl a fhios agam!!!!

[@ddorran]

Níl a fhios agam!!!!

[@JuanAntonio-ff4mg]

" NIl a fhion agam",... No te entiendo compadre, hablame en cristiano xd

[@ddorran]

porque diablos no habla irlandesa :)