divided an audio signal into no. of frames (segments) and Compute mean from each frame (segment) and then how to Concatenate each mean of a frame to obtain one feature vector

[y, Fs] = audioread('F:\audio samples\car crashes\(resource)car_crash.wav');
N = length(y); % sample length
slength = N/Fs; % total time span of audio signal
t = linspace(0, N/Fs, N);
figure;
subplot(2,2,1);
plot(t, y); % pplots the audio
duration = round(0.04*Fs); %duration selected from  40 ms, how many samples in a frame
n_f=floor(N/Fs)+1; %how many frames are there 
display(n_f);
temp=0; %temporary value
for i= 1: n_f
frames(i,:) =y (temp+1: temp+duration);
temp=temp+duration;
M{i} =mean(frames(i,:));%mean of (n_f) frames
display(M{i});
Mconcat = cat(1,M{i});
display(Mconcat);

[Walter Roberson]

The title does not ask for 40 ms, it asks for 40 total frames.

The Mconcat you create should be the vector of means. You could eliminate some of the trouble by using

Mconcat = zeros(1,n_f);
for i = 1: n_f
  frames(i,:) =y (temp+1: temp+duration);
  temp=temp+duration;
  Mconcat(i) = mean(frames(i,:));%mean of (n_f) frames
end