TAMA

Frame

class entrainment_metrics.tama.frame.Frame(start: float, end: float, is_missing: bool, interpausal_units: Optional[List[InterPausalUnit]])

An interval of time inside an audio

…

start

Start time of the frame.

Type

float

end

End time of the frame.

Type

float

is_missing

Whether the frame has no IPUs inside. In other words, if the frame is fulled with silence.

Type

bool

interpausal_units

The IPUs that fall inside of the frame

Type

List[InterPausalUnit]

calculate_feature_value(feature: str, audio_file: Optional[Path] = None, pitch_gender: Optional[str] = None, extractor: Optional[str] = None) → float

Return the frame’s value for the feature given

This value is calculated as the duration-weighted mean of the value for the feature of each IPU inside the frame

Cite Interspeech2016

class entrainment_metrics.tama.frame.MissingFrame(start: float, end: float)

calculate_feature_value(feature: str, audio_file: Optional[Path] = None, pitch_gender: Optional[str] = None, extractor: Optional[str] = None) → float

Return the frame’s value for the feature given

This value is calculated as the duration-weighted mean of the value for the feature of each IPU inside the frame

Cite Interspeech2016

Utils

entrainment_metrics.tama.utils.get_frames(wav_fname: Path, words_fname: Path) → List[Union[Frame, MissingFrame]]

Return a list of Frames given a Path to a .word file and a .wav file

The format of the word file must be: - For each line f’{starting_time} {ending_time} {word}’ Where starting_time and ending_time are floats

Parameters

• wav_fname (Path) – The path to the wav file

• words_fname (Path) – The path to the words file

Returns

The frames from the wav file with the InterPausalUnits from the word file.

Return type

List[Union[Frame, MissingFrame]]

Entrainment

entrainment_metrics.tama.entrainment.calculate_sample_correlation(time_series_a: List[float], time_series_b: List[float], lags: int) → List[float]

Calculate the correlations between two series as one of them is lagged

Intuitively, it can be interpreted similarly to Pearson’s correlation coeffi- cient between a time-series and a lagged version of another one, which means that its value varies from −1 to 1. Each value returned can be interpreted as an indication of how much a speaker converged (diverged) in a task in terms of the behavior of a/p feature φ to the behavior her partner had h frames before, where h is the number of lags.

entrainment_metrics.tama.entrainment.calculate_time_series(feature: str, frames: List[Union[Frame, MissingFrame]], audio_file: Optional[Path] = None, extractor: Optional[str] = None, pitch_gender: Optional[str] = None) → List[float]

Generate a time series of the frames values for the feature given