In filipezabala/voice: Voice Analysis, Speaker Recognition and Mood Inference via music theory

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
ini0 <- Sys.time()

0. Installation

0.1 Minimal

# CRAN (stable)
install.packages('voice', dep = TRUE)

# Github (development)
devtools::install_github('filipezabala/voice')

0.2 Full

https://github.com/filipezabala/voice

1. Extract features

1.1 Get path to audio files

wavFiles <- list.files(system.file('extdata', package = 'wrassp'),
                       pattern <- glob2rx('*.wav'), full.names = TRUE)

1.2 Minimal usage

M <- voice::extract_features(wavFiles)
dplyr::glimpse(M)

2. Tag

2.1 Creating synthetic Extended data

E <- dplyr::tibble(subject_id = c(1,1,1,2,2,2,3,3,3), wav_path = wavFiles)
E

2.2 Tagging Extended data

voice::tag(E)

2.3 Tagging Canonical data

voice::tag(E, groupBy = 'subject_id')

3. Voice2Sheet (experimental)

3.1 Get audio

url0 <- 'https://github.com/filipezabala/voiceAudios/blob/main/wav/doremi.wav?raw=true'
download.file(url0, paste0(tempdir(), '/doremi.wav'), mode = 'wb')
# embedr::embed_audio(url0)

3.2 Extract F0

M <- voice::extract_features(tempdir(), features = 'f0')
summary(M)

3.3 Plot

# Plot
plot(M$f0)
legend(-60, 170, 'Do (C3)' , bty = 'n')
legend(80, 190 , 'Re (D3)' , bty = 'n')
legend(220, 290, 'Mi (E4)' , bty = 'n')
legend(350, 210, 'Fa (F3)' , bty = 'n')
legend(480, 250, 'Sol (G3)', bty = 'n')
legend(600, 270, 'La (A3)' , bty = 'n')
legend(720, 310, 'Si (B3)' , bty = 'n')
legend(910, 320, 'Do (C4)' , bty = 'n')

3.4 Assign notes (manually)

(note1 <- median(M$f0[1:150], na.rm = TRUE))
voice::notes(note1)
(note2 <- median(M$f0[190:280], na.rm = TRUE))
voice::notes(note2)
(note3 <- median(M$f0[310:380], na.rm = TRUE))
voice::notes(note3)
(note4 <- median(M$f0[420:500], na.rm = TRUE))
voice::notes(note4)
(note5 <- median(M$f0[590:630], na.rm = TRUE))
voice::notes(note5)
(note6 <- median(M$f0[700:770], na.rm = TRUE))
voice::notes(note6)
(note7 <- median(M$f0[820:900], na.rm = TRUE))
voice::notes(note7)
(note8 <- median(M$f0[950:1100], na.rm = TRUE))
voice::notes(note8)

3.5 Music sheet

# gather (manually)
(notes_vec <- c(paste0('note', 1:8), 'note8'))
(f0_spn <- voice::notes(mget(notes_vec)))

# duration
(dur <- voice::duration(f0_spn))

# gm by Renfei Mao
m <- gm::Music()
m <- m +
  gm::Meter(4, 4) +
  gm::Line(pitches = as.list(as.character(dur$note)),
           durations = as.list(dur$dur_line)) +
  gm::Tempo(100)
# gm::show(m, to = c('score', 'audio'))

filipezabala/voice documentation built on April 12, 2025, 12:39 p.m.