# Song analysis shared methods

# Song analysis software packages

# Matlab functions

Function |
Author |
Description |

extract_features | Sigal Saar | Calculates spectral derivatives and acoustic features (such as mean amplitude modulation, mean frequency modulation, mean Wiener Entropy, mean amplitude, "gravity center" which is mean mean frequency, mean goodness of pitch, mean pitch) from WAVE data. .wav file is read into matlab using the "wavread" command. |

crosscorrelate | Primoz Ravbar |
Cross-correlates song feature vectors 'a' and 'b'; ('a' and 'b' are the inputs) |

local_realignment | Primoz Ravbar |
The function locally realigns rows in a matrix according to maximal crosscorrelations between the locally selected frames. The algorithm tracks locally identified frames from the last row to the firs row in the matrix. The function was written for the purpose of realigning acoustic features of syllables in zebra finch song. Each row of the input data was an acoustic feature, sampled with millisecond resolution. |

track_vocal_changes | Primoz Ravbar |
The script was originally developed to track significant vocal changes across the developing syllable, where syllables were rows in the matrix (called 'm_tot_pm2') and columns corresponded to milliseconds. It could be used to identify any significant changes that occur over a user-specified frame (number of rows in any particular millisecond). To identify the changes, all rows in the matrix must be locally realigned. The script calls function 'local_realignment' to perform local realignment using best cross-correlation method (see 'local_realignment' comments). |

smooth2_function | Primoz Ravbar |
The function performs 2-D smoothing of a matrix using Hanning window of user-specified size: window_x=columns; window_y=rows; if window_x or window_y equlas 1 no smoothing will be performed;inputs: matrix of data to be smothened -- 's_tot'; size of Hanning window for x axis -- 'window_x'; size of Hanning window for y axis -- 'window_y'; output: smoothened matrix |

K-L Distance Analysis | Richard Bertram |
Description: Used to compare two song motifs using the Kullback-Leibler (K-L) distance between the two-dimensional scatter plots of features. These features are computed using the Sound Analysis Pro software. Full description in Journal of Neuroscience Methods, 174:147, 2008. |