utils.py

# utils.py
import numpy as np

# Function for computing mean and std for calibration
def personal_baseline(f0, rms, frame_times, speech_mask, T_cal):
    cal_mask = (frame_times <= T_cal) & speech_mask

    # pitch baseline (voiced only)
    voiced_cal = cal_mask & (~np.isnan(f0))
    if np.sum(voiced_cal) >= 8:
        mu_f0 = float(np.mean(f0[voiced_cal]))
        sd_f0 = float(np.std(f0[voiced_cal]) + 1e-8)
    else:
        mu_f0, sd_f0 = float(np.nanmean(f0)), float(np.nanstd(f0) + 1e-8)

    # energy baseline (speech only)
    if np.sum(cal_mask) >= 8:
        mu_e = float(np.mean(rms[cal_mask]))
        sd_e = float(np.std(rms[cal_mask]) + 1e-8)
    else:
        mu_e, sd_e = float(np.mean(rms)), float(np.std(rms) + 1e-8)

    return mu_f0, sd_f0, mu_e, sd_e