source: main/trunk/model-sites-dev/mars/collect/deam/essentia-2013-2014.profile@ 34430

Last change on this file since 34430 was 34430, checked in by davidb, 4 years ago

Used to extract audio features from DEAM MP3 files from the 2013 and 2014 years of collectcion (45 secs each)

File size: 1.2 KB
Line 
1# The Deam 2013-2014 dataset are all 45 second samples
2# Looking to match with up with Grekow's results with Weka
3# using his private dataset where users rated Arousal
4# and Valence after listening to 6 seconds of music
5
6# => compute features for 6 seconds of DEAM audio starting 25 seconds in
7startTime: 25
8endTime: 31
9
10outputFrames: 1
11outputFormat: json
12
13#requireMbid: false
14
15# Standardize on 44.1 kHz as a sample rate
16analysisSampleRate: 44100.0
17
18lowlevel:
19# 2048 and 1024 are the extractors reported default values
20# frameSize: 2048
21# hopSize: 1024
22
23# Considered working with a slightly bigger framesize
24# frameSize: 4096
25# hopSize: 2048
26
27# Ultimately decided to work with slightly bigger framesize
28# align with 100msec intervales to better mesh with DEAM
29# ground-truth data
30#
31# => At 44.1 kHz analysisSampleRate, a hop size of 2205 produces
32# a frame of data for every 0.1 sec of audio
33 frameSize: 4096
34 hopSize: 2205
35
36
37# The following are some example values from the Essentia web page on extractor CLI tool
38# zeroPadding: 0
39# windowType: blackmanharris62
40# silentFrames: noise
41# stats: ["mean", "var", "median", "min", "max", "dmean", "dmean2", "dvar", "dvar2"]
Note: See TracBrowser for help on using the repository browser.