timit

Documentation

Overview

Package timit Phones of the TIMIT database. For recognition testing the full set of 61 is typically reduced to 39 with confusable sounds folded into a group, e.g. "sh" and "zh" See Speaker-Independent Phone Recognition Using Hidden Markov Models, Kai-Fu Lee and Hsiao-Wuen Hon in IEEE Transactions on Acoustics, Speech and Signal Processing, Vol 37, 1989 for the original set and collapsing to 39 phones Many later studies use the 39 phone set

Index

• Variables
• func IdxFmSnd(s string, id string) (v int, ok bool)
• func IdxFmSnd2(s string, id string) (v int, ok bool)
• func LoadText(fn string) (string, error)
• func LoadTimes(fn string, names []string) ([]speech.Unit, error)
• func LoadTranscription(fn string) ([]string, error)
• func SndFmIdx(idx int, id string) (phone string, ok bool)
• func SndFmIdx2(idx int, id string) (phone string, ok bool)

Variables

var PhoneCats = []string{"iy", "ih", "eh", "ae", "ix", "ah", "uw", "uh", "ao", "ey",
	"ay", "oy", "aw", "ow", "l", "r", "y", "w", "er", "m", "n", "ng", "ch", "jh", "dh", "b", "d", "dx",
	"g", "p", "t", "k", "z", "zh", "v", "f", "th", "s", "hh", "pcl", "q"}

PhoneCats and PhoneMap must maintain same order!

var PhoneList = []string{"iy", "ih", "eh", "ae", "ix", "ah", "ax", "ax-h", "uw", "ux", "uh", "ao", "aa", "ey",
	"ay", "oy", "aw", "ow", "l", "el", "r", "y", "w", "er", "axr", "m", "em", "n", "nx", "en", "ng",
	"eng", "ch", "jh", "dh", "b", "d", "dx", "g", "p", "t", "k", "z", "zh", "v", "f", "th", "s", "sh",
	"hh", "hv", "pcl", "tcl", "kcl", "bcl", "dcl", "gcl", "epi", "h#", "pau", "q"}

PhoneList is the full list of phones. Some phones get folded together and the reduced set is the PhoneCats variable.

var Phones = map[string]int{
	"iy":   0,
	"ih":   1,
	"eh":   2,
	"ae":   3,
	"ix":   4,
	"ah":   5,
	"ax":   5,
	"ax-h": 5,
	"uw":   6,
	"ux":   6,
	"uh":   7,
	"ao":   8,
	"aa":   8,
	"ey":   9,
	"ay":   10,
	"oy":   11,
	"aw":   12,
	"ow":   13,
	"l":    14,
	"el":   14,
	"r":    15,
	"y":    16,
	"w":    17,
	"er":   18,
	"axr":  18,
	"m":    19,
	"em":   19,
	"n":    20,
	"nx":   20,
	"en":   20,
	"ng":   21,
	"eng":  21,
	"ch":   22,
	"jh":   23,
	"dh":   24,
	"b":    25,
	"d":    26,
	"dx":   27,
	"g":    28,
	"p":    29,
	"t":    30,
	"k":    31,
	"z":    32,
	"zh":   33,
	"sh":   33,
	"v":    34,
	"f":    35,
	"th":   36,
	"s":    37,
	"hh":   38,
	"hv":   38,
	"pcl":  39,
	"tcl":  39,
	"kcl":  39,
	"bcl":  39,
	"dcl":  39,
	"gcl":  39,
	"h#":   39,
	"pau":  39,
	"epi":  39,
	"q":    40,
}

var Phones2 = map[string]int{
	"iy":   0,
	"ih":   1,
	"eh":   2,
	"ae":   3,
	"ix":   4,
	"ah":   5,
	"ax":   5,
	"ax-h": 5,
	"uw":   6,
	"ux":   6,

	"ao": 7,
	"aa": 7,
	"ey": 8,
	"ay": 9,

	"ow":  10,
	"l":   11,
	"el":  11,
	"r":   12,
	"y":   13,
	"w":   14,
	"er":  15,
	"axr": 15,
	"m":   16,
	"em":  16,
	"n":   17,
	"nx":  17,
	"en":  17,

	"ch": 18,
	"jh": 19,
	"dh": 20,
	"b":  21,
	"d":  22,
	"dx": 23,

	"p":   24,
	"t":   25,
	"k":   26,
	"z":   27,
	"zh":  28,
	"sh":  28,
	"v":   29,
	"f":   30,
	"th":  31,
	"s":   32,
	"hh":  33,
	"hv":  33,
	"pcl": 34,
	"tcl": 34,
	"kcl": 34,
	"bcl": 34,
	"dcl": 34,
	"gcl": 34,
	"h#":  34,
	"pau": 34,
	"epi": 34,
	"q":   35,
}

Functions

func IdxFmSnd

func IdxFmSnd(s string, id string) (v int, ok bool)

IdxFmSnd returns the slice index of the snd if found. id is ignored if the corpus doesn't have subsets of sounds

func IdxFmSnd2

func IdxFmSnd2(s string, id string) (v int, ok bool)

IdxFmSnd returns the slice index of the snd if found. id is ignored if the corpus doesn't have subsets of sounds

func LoadText

func LoadText(fn string) (string, error)

LoadText retrieves the full text of the timit transcription

func LoadTimes

func LoadTimes(fn string, names []string) ([]speech.Unit, error)

LoadTimes loads both the timing and transcription data for timit files so the names slice is unused

func LoadTranscription

func LoadTranscription(fn string) ([]string, error)

LoadTranscription is a "no op" for timit, LoadTimes does the work of both

func SndFmIdx

func SndFmIdx(idx int, id string) (phone string, ok bool)

SndFmIdx returns the sound if found in the map of sounds of the corpus. id is ignored if the corpus doesn't have subsets of sounds

func SndFmIdx2

func SndFmIdx2(idx int, id string) (phone string, ok bool)

SndFmIdx returns the sound if found in the map of sounds of the corpus. id is ignored if the corpus doesn't have subsets of sounds