(1) Plot the power spectral densities of b0002.wav and b0018.wav as a function of real frequency. (The
sampling frequency is 2 KHz.) For each of the two data files, what is the frequency that has the
maximum power spectral density?
(2) With all the training data, run the following three models: medium KNN, medium tree, and medium
Gaussian SVM. What is the overall accuracy of each model? What is the rate of false-negative rate
for each model? Report the confusion matrix for each model. What is the size of your feature _table?
(3) In table variable feature _table, combine the median Value and mean Value to make
mm Value=(median Value +mean Value)/2; combine standard Deviation and mean Absolute Deviation
to make Deviation=sqrt(standard Deviation*mean Absolute Deviation).
Use mm Value and Deviation
as new features to replace median Value, mean Value, standard Deviation, and
mean Absolute Deviation to create a new feature table (feature_table2). Redo Part (2) to with
feature_table2.
Requirements: 200 | .doc file
