soundgen (version 2.7.2)

getSmoothContour: Smooth contour from anchors

Description

Returns a smooth contour based on an arbitrary number of anchors. Used by soundgen for generating intonation contour, mouth opening, etc. This function is mostly intended to be used internally by soundgen, more precisely to construct (upsample) smooth curves from a number of anchors. For general upsampling or downsampling of audio, use resample. Note that pitch contours are treated as a special case: values are log-transformed prior to smoothing, so that with 2 anchors we get a linear transition on a log scale (as if we were operating with musical notes rather than frequencies in Hz). Pitch plots have two Y axes: one showing Hz and the other showing musical notation.

Usage

getSmoothContour(
  anchors = data.frame(time = c(0, 1), value = c(0, 1)),
  len = NULL,
  thisIsPitch = FALSE,
  normalizeTime = TRUE,
  interpol = c("approx", "spline", "loess")[3],
  loessSpan = NULL,
  discontThres = 0.05,
  jumpThres = 0.01,
  valueFloor = NULL,
  valueCeiling = NULL,
  plot = FALSE,
  xlim = NULL,
  ylim = NULL,
  xlab = "Time, ms",
  ylab = ifelse(thisIsPitch, "Frequency, Hz", "Amplitude"),
  main = ifelse(thisIsPitch, "Pitch contour", ""),
  samplingRate = 16000,
  voiced = NULL,
  contourLabel = NULL,
  NA_to_zero = TRUE,
  ...
)

Value

Returns a numeric vector of length len.

Arguments

anchors

a numeric vector of values or a list/dataframe with one column (value) or two columns (time and value). achors$time can be in ms (with len=NULL) or in arbitrary units, eg 0 to 1 (with duration determined by len, which must then be provided in ms). So anchors$time is assumed to be in ms if len=NULL and relative if len is specified. anchors$value can be on any scale.

len

the required length of the output contour. If NULL, it will be calculated based on the maximum time value (in ms) and samplingRate

thisIsPitch

(boolean) is this a pitch contour? If true, log-transforms before smoothing and plots in both Hz and musical notation

normalizeTime

if TRUE, normalizes anchors$time values to range from 0 to 1

interpol

method of interpolation between anchors: "approx" = linear with approx, "spline" = cubic splines with spline, "loess" = local polynomial regression with loess

loessSpan

controls the amount of smoothing when interpolating between anchors with loess, so only has an effect if interpol = 'loess' (1 = strong, 0.5 = weak smoothing)

discontThres

if two anchors are closer in time than discontThres (on a 0-1 scale, ie specified as proportion of total length), the contour is broken into segments with a linear transition between these segments

jumpThres

if anchors are closer than jumpThres, a new section starts with no transition at all (e.g. for adding pitch jumps)

valueFloor, valueCeiling

lowser/upper bounds for the contour

plot

(boolean) produce a plot?

xlim, ylim, xlab, ylab, main

plotting options

samplingRate

sampling rate used to convert time values to points (Hz)

voiced, contourLabel

graphical pars for plotting breathing contours (see examples below)

NA_to_zero

if TRUE, all NAs are replaced with zero

...

other plotting options passed to plot()

Examples

# long format: anchors are a dataframe
a = getSmoothContour(anchors = data.frame(
  time = c(50, 137, 300), value = c(0.03, 0.78, 0.5)),
  normalizeTime = FALSE,
  voiced = 200, valueFloor = 0, plot = TRUE, main = '',
  samplingRate = 16000) # breathing

# short format: anchors are a vector (equal time steps assumed)
a = getSmoothContour(anchors = c(350, 800, 600),
  len = 5500, thisIsPitch = TRUE, plot = TRUE,
  samplingRate = 3500) # pitch

# a single anchor gives constant value
a = getSmoothContour(anchors = 800,
  len = 500, thisIsPitch = TRUE, plot = TRUE, samplingRate = 500)

# two pitch anchors give loglinear F0 change
a = getSmoothContour(anchors = c(220, 440),
  len = 500, thisIsPitch = TRUE, plot = TRUE, samplingRate = 500)

## Two closely spaced anchors produce a pitch jump
# one loess for the entire contour
a1 = getSmoothContour(anchors = list(time = c(0, .15, .2, .7, 1),
    value = c(360, 116, 550, 700, 610)), len = 500, thisIsPitch = TRUE,
    plot = TRUE, samplingRate = 500)
# two segments with a linear transition
a2 = getSmoothContour(anchors = list(time = c(0, .15, .17, .7, 1),
    value = c(360, 116, 550, 700, 610)), len = 500, thisIsPitch = TRUE,
    plot = TRUE, samplingRate = 500)
# two segments with an abrupt jump
a3 = getSmoothContour(anchors = list(time = c(0, .15, .155, .7, 1),
    value = c(360, 116, 550, 700, 610)), len = 500, thisIsPitch = TRUE,
    plot = TRUE, samplingRate = 500)
# compare:
plot(a2)
plot(a3)  # NB: the segment before the jump is upsampled to compensate

## Control the amount of smoothing
getSmoothContour(c(1, 3, 9, 10, 9, 9, 2), len = 100, plot = TRUE,
  loessSpan = NULL)  # default amount of smoothing (depends on dur)
getSmoothContour(c(1, 3, 9, 10, 9, 9, 2), len = 100, plot = TRUE,
  loessSpan = .85)   # more smoothing than default
getSmoothContour(c(1, 3, 9, 10, 9, 9, 2), len = 100, plot = TRUE,
  loessSpan = .5)    # less smoothing
getSmoothContour(c(1, 3, 9, 10, 9, 9, 2), len = 100, plot = TRUE,
  interpol = 'approx')  # linear interpolation (no smoothing)

## Upsample preserving leading and trailing NAs
anchors = data.frame(time =  c(1,  4,  5,  7,  10, 20, 23, 25, 30),
                     value = c(NA, NA, 10, 15, 12, NA, 17, 15, NA))
plot(anchors, type = 'b')
anchors_ups = getSmoothContour(
  anchors, len = 200,
  interpol = 'approx',  # only approx can propagate NAs
  NA_to_zero = FALSE,   # preserve NAs
  discontThres = 0)     # don't break into sub-contours
plot(anchors_ups, type = 'b')