clarity.evaluator.hasqi package

Submodules

clarity.evaluator.hasqi.hasqi module

clarity.evaluator.hasqi.hasqi.hasqi_v2(reference: ndarray, reference_sample_rate: float, processed: ndarray, processed_sample_rate: float, audiogram: Audiogram, equalisation: int = 1, level1: float = 65.0, silence_threshold: float = 2.5, add_noise: float = 0.0, segment_covariance: int = 16) → tuple[float, float, float, list[float]][source]

Function to compute the HASQI version 2 quality index using the auditory model followed by computing the envelope cepstral correlation and BM vibration average short-time coherence signals. The reference signal presentation level for NH listeners is assumed to be 65 dB SPL. The same model is used for both normal and impaired hearing.

Parameters:

  • reference (np.ndarray) – Clear input reference speech signal with no noise or distortion. If a hearing loss is specified, NAL-R equalization is optional

  • reference_sample_Rate (int) – Sampling rate in Hz for reference signal.

  • processed (np.ndarray) – Output signal with noise, distortion, HA gain, and/or processing.

  • processed_sample_rate (int) – Sampling rate in Hz for processed signal.

  • hearing_loss (np.ndarray) – vector of hearing loss at the 6 audiometric frequencies [250, 500, 1000, 2000, 4000, 6000] Hz.

  • equalisation (int) – Mode to use when equalising the reference signal: 1 = no EQ has been provided, the function will add NAL-R 2 = NAL-R EQ has already been added to the reference signal

  • level1 – Optional input specifying level in dB SPL that corresponds to a signal RMS = 1. Default is 65 dB SPL if argument not provided.

  • silence_threshold (float) – Silence threshold sum across bands, dB above audio threshold. Default: 2.5

  • add_noise (float) – Additive noise in dB SL to conditional cross-covariance. Default is 0.0

  • segment_covariance (int) – Segment size for the covariance calculation. Default is 16

Returns:

tuple(Combined, Nonlin, Linear, raw)

Combined: Quality estimate is the product of the nonlinear and linear terms Nonlin: Nonlinear quality component = (cepstral corr)^2 x seg BM coherence Linear: Linear quality component = std of spectrum and spectrum slope raw: Vector of raw values = [CepCorr, BMsync5, Dloud, Dslope]

James M. Kates, 5 August 2013. Translated from MATLAB to Python by Gerardo Roa Dabike, October 2022.

clarity.evaluator.hasqi.hasqi.hasqi_v2_better_ear(reference_left: ndarray, reference_right: ndarray, processed_left: ndarray, processed_right: ndarray, sample_rate: float, listener: Listener, level: float = 100.0) → float[source]

Better ear HASQI.

Calculates HASQI for left and right ear and selects the better result.

Parameters:

  • reference_left (np.ndarray) – left channel of reference signal

  • reference_right (np.ndarray) – right channel of reference signal

  • reference_left – left channel of processed signal

  • reference_right – right channel of processed signal

  • sample_rate – sampling rate for both signal

  • audiogram_l – left ear audiogram

  • audiogram_r – right ear audiogram

  • level – level in dB SPL corresponding to RMS=1

  • audiogram_freq – selected frequencies to use for audiogram

Returns:

beHASQI score

Return type:

float

Gerardo Roa Dabike, November 2022

Module contents

clarity.evaluator.hasqi.hasqi_v2(reference: ndarray, reference_sample_rate: float, processed: ndarray, processed_sample_rate: float, audiogram: Audiogram, equalisation: int = 1, level1: float = 65.0, silence_threshold: float = 2.5, add_noise: float = 0.0, segment_covariance: int = 16) → tuple[float, float, float, list[float]][source]

Function to compute the HASQI version 2 quality index using the auditory model followed by computing the envelope cepstral correlation and BM vibration average short-time coherence signals. The reference signal presentation level for NH listeners is assumed to be 65 dB SPL. The same model is used for both normal and impaired hearing.

Parameters:

  • reference (np.ndarray) – Clear input reference speech signal with no noise or distortion. If a hearing loss is specified, NAL-R equalization is optional

  • reference_sample_Rate (int) – Sampling rate in Hz for reference signal.

  • processed (np.ndarray) – Output signal with noise, distortion, HA gain, and/or processing.

  • processed_sample_rate (int) – Sampling rate in Hz for processed signal.

  • hearing_loss (np.ndarray) – vector of hearing loss at the 6 audiometric frequencies [250, 500, 1000, 2000, 4000, 6000] Hz.

  • equalisation (int) – Mode to use when equalising the reference signal: 1 = no EQ has been provided, the function will add NAL-R 2 = NAL-R EQ has already been added to the reference signal

  • level1 – Optional input specifying level in dB SPL that corresponds to a signal RMS = 1. Default is 65 dB SPL if argument not provided.

  • silence_threshold (float) – Silence threshold sum across bands, dB above audio threshold. Default: 2.5

  • add_noise (float) – Additive noise in dB SL to conditional cross-covariance. Default is 0.0

  • segment_covariance (int) – Segment size for the covariance calculation. Default is 16

Returns:

tuple(Combined, Nonlin, Linear, raw)

Combined: Quality estimate is the product of the nonlinear and linear terms Nonlin: Nonlinear quality component = (cepstral corr)^2 x seg BM coherence Linear: Linear quality component = std of spectrum and spectrum slope raw: Vector of raw values = [CepCorr, BMsync5, Dloud, Dslope]

James M. Kates, 5 August 2013. Translated from MATLAB to Python by Gerardo Roa Dabike, October 2022.

clarity.evaluator.hasqi.hasqi_v2_better_ear(reference_left: ndarray, reference_right: ndarray, processed_left: ndarray, processed_right: ndarray, sample_rate: float, listener: Listener, level: float = 100.0) → float[source]

Better ear HASQI.

Calculates HASQI for left and right ear and selects the better result.

Parameters:

  • reference_left (np.ndarray) – left channel of reference signal

  • reference_right (np.ndarray) – right channel of reference signal

  • reference_left – left channel of processed signal

  • reference_right – right channel of processed signal

  • sample_rate – sampling rate for both signal

  • audiogram_l – left ear audiogram

  • audiogram_r – right ear audiogram

  • level – level in dB SPL corresponding to RMS=1

  • audiogram_freq – selected frequencies to use for audiogram

Returns:

beHASQI score

Return type:

float

Gerardo Roa Dabike, November 2022