MATLAB中采用Savitzky-Golay低通滤波器处理心跳录音信号的方法

Hey there! Let's break down exactly how to use MATLAB's Savitzky-Golay filter on your loaded heartbeat audio data, including parameter setup and proper function syntax.

First: Quick Recap of Your Setup #

You already have your heartbeat audio loaded into a MATLAB variable (let's assume you used audioread, so your signal is stored in y with sampling rate Fs). If your audio is stereo, make sure to convert it to mono first (we'll include that in the example below).

Savitzky-Golay Filter Parameters: What to Choose? #

The Savitzky-Golay filter relies on two key parameters—here's how to pick them for heartbeat signals:

• Polynomial Order: This controls how well the filter fits local signal trends. For heartbeat data (which has smooth, periodic patterns), stick to 2 or 3. Higher orders can overfit to noise, while lower orders might oversimplify the signal.
• Window Length: Must be an odd integer (since it centers on each data point). For typical audio sampling rates (e.g., 44100 Hz or 48000 Hz), start with values between 7 and 15.
  • Larger window lengths = stronger smoothing (great for reducing high-frequency noise like microphone hiss)
  • Smaller window lengths = preserves more signal detail (use if you notice the filtered signal is losing sharp heartbeat peaks)

How to Call the Filter (Where to Put Your WAV Variable) #

MATLAB's sgolayfilt function is the tool you need, and your loaded WAV signal variable goes as the first input argument. Here's a complete example:

% 1. Load your WAV data (you've already done this, included for context)
[y, Fs] = audioread('your_heartbeat_file.wav');

% Convert stereo to mono if needed
if size(y, 2) > 1
    y = y(:, 1); % Grab left channel; adjust to (:,2) for right if preferred
end

% 2. Set your Savitzky-Golay parameters
poly_order = 2;       % Start with 2, tweak if needed
window_length = 11;   % Odd integer, start with 11

% 3. Apply the filter: your WAV variable 'y' is the first argument
y_filtered = sgolayfilt(y, poly_order, window_length);

Tuning Tips #

• If the filtered signal still has too much high-frequency noise: Increase window_length (try 13, 15)
• If the heartbeat peaks look blurred/lost: Decrease window_length (try 7,9) or lower the polynomial order to 1
• To visualize the difference between original and filtered signals, add this code:

  t = (0:length(y)-1)/Fs; % Time vector for plotting
  figure;
  subplot(2,1,1);
  plot(t, y);
  title('Original Heartbeat Signal');
  xlabel('Time (s)'); ylabel('Amplitude');
  subplot(2,1,2);
  plot(t, y_filtered);
  title('Filtered Heartbeat (Savitzky-Golay)');
  xlabel('Time (s)'); ylabel('Amplitude');
  

内容的提问来源于stack exchange，提问作者Mayur Makhija