import numpy as np from scipy.optimize import least_squares from scipy.optimize import minimize from scipy.linalg import lstsq as sclstsq import scipy.linalg as lin def harmonicEuler ( sN, fs, t, f0, k1, kN, ks ): """ Performs inverse calculation of harmonics contaminating a signal. Args: sN = signal containing noise fs = sampling frequency t = time samples f0 = base frequency of the sinusoidal noise nK = number of harmonics to calculate """ KK = np.arange(k1, kN+1, 1/ks ) nK = len(KK) A = np.exp(1j* np.tile(KK,(len(t), 1)) * 2*np.pi* (f0/fs) * np.tile(np.arange(1, len(t)+1, 1),(nK,1)).T) v = np.linalg.lstsq(A, sN, rcond=None) alpha = np.real(v[0]) beta = np.imag(v[0]) amp = np.abs(v[0]) phase = np.angle(v[0]) h = np.zeros(len(t)) for ik, k in enumerate(KK): h += 2*amp[ik] * np.cos( 2.*np.pi*(k) * (f0/fs) * np.arange(1, len(t)+1, 1 ) + phase[ik] ) return sN-h def harmonicNorm (f0, sN, fs, t, k1, kN, ks): #return np.linalg.norm( harmonicEuler(sN, fs, t, f0, k1, kN, ks)) ii = sN < (3.* np.std(sN)) return np.linalg.norm( harmonicEuler(sN, fs, t, f0, k1, kN, ks)[ii] ) def minHarmonic(sN, fs, t, f0, k1, kN, ks): # CG, BFGS, Newton-CG, L-BFGS-B, TNC, SLSQP, dogleg, trust-ncg, trust-krylov, trust-exact and trust-constr #res = minimize(harmonicNorm, np.array((f0)), args=(sN, fs, t, k1, kN, ks), jac='2-point', method='BFGS') # hess=None, bounds=None ) ############# # Reduced N # ############# #print("min 10") #res = minimize(harmonicNorm, np.array((f0)), args=(sN, fs, t, k1, 10, ks), jac='2-point', method='BFGS') # hess=None, bounds=None ) ########## # Bounds # ########## print("bounded search for f0", f0) bnds = ( (f0-0.125, f0+0.125), ) #print("len f0", len( [f0,] )) #print("len bnds", len(bnds) ) res = minimize(harmonicNorm, (f0,), args=(sN, fs, t, k1, kN, ks), jac='2-point', method='L-BFGS-B', bounds=bnds ) # hess=None, bounds=None ) #print(res) #print ( "guess", guessf0( harmonicEuler(sN, fs, t, res.x[0], k1, kN, ks), fs ) ) return harmonicEuler(sN, fs, t, res.x[0], k1, kN, ks), res.x[0]#[0] def harmonicEuler2 ( sN, fs, t, f0, f0k1, f0kN, f0ks, f1, f1k1, f1kN, f1ks ): """ Performs inverse calculation of harmonics contaminating a signal. Args: sN = signal containing noise fs = sampling frequency t = time samples f0 = first base frequency of the sinusoidal noise f0k1 = First harmonic to calulate for f0 f0kN = Last base harmonic to calulate for f0 f0ks = subharmonics to calculate f1 = second base frequency of the sinusoidal noise f1k1 = First harmonic to calulate for f1 f1kN = Last base harmonic to calulate for f1 f1ks = subharmonics to calculate at f1 base frequency """ KK0 = np.arange(f0k1, f0kN+1, 1/f0ks) nK0 = len(KK0) A0 = np.exp(1j* np.tile(KK0,(len(t), 1)) * 2*np.pi* (f0/fs) * np.tile( np.arange(1, len(t)+1, 1), (nK0,1)).T) KK1 = np.arange(f1k1, f1kN+1, 1/f1ks) nK1 = len(KK1) A1 = np.exp(1j* np.tile(KK1,(len(t), 1)) * 2*np.pi* (f1/fs) * np.tile( np.arange(1, len(t)+1, 1),(nK1,1)).T) A = np.concatenate((A0, A1), axis=1) v = np.linalg.lstsq(A, sN, rcond=None) # rcond=None) #, rcond=1e-8) amp0 = np.abs(v[0][0:nK0]) phase0 = np.angle(v[0][0:nK0]) amp1 = np.abs(v[0][nK0::]) phase1 = np.angle(v[0][nK0::]) h = np.zeros(len(t)) for ik, k in enumerate(KK0): h += 2*amp0[ik] * np.cos( 2.*np.pi*(k) * (f0/fs) * np.arange(1, len(t)+1, 1 ) + phase0[ik] ) for ik, k in enumerate(KK1): h += 2*amp1[ik] * np.cos( 2.*np.pi*(k) * (f0/fs) * np.arange(1, len(t)+1, 1 ) + phase1[ik] ) return sN-h def harmonic2Norm (f0, sN, fs, t, f0k1, f0kN, f0ks, f1k1, f1kN, f1ks): #return np.linalg.norm(harmonicEuler2(f0[0], f0[1], sN, fs, nK, t)) ii = sN < (3.* np.std(sN)) return np.linalg.norm( harmonicEuler2(sN, fs, t, f0[0], f0k1, f0kN, f0ks, f0[1], f1k1, f1kN, f1ks)[ii] ) def minHarmonic2(sN, fs, t, f0, f0k1, f0kN, f0ks, f1, f1k1, f1kN, f1ks): # CG, BFGS, Newton-CG, L-BFGS-B, TNC, SLSQP, dogleg, trust-ncg, trust-krylov, trust-exact and trust-constr #res = minimize(harmonic2Norm, np.array((f0, f1)), args=(sN, fs, t, f0k1, f0kN, f0ks, f1k1,f1kN, f1ks), jac='2-point', method='BFGS') # hess=None, bounds=None ) print("search 10") res = minimize(harmonic2Norm, np.array((f0, f1)), args=(sN, fs, t, f0k1, 10, f0ks, f1k1, 10, f1ks), jac='2-point', method='BFGS') # hess=None, bounds=None ) #print(res) #print ( "guess", guessf0(harmonicEuler2(sN, fs, t, res.x[0], f0k1, f0kN, f0ks, res.x[1], f1k1, f1kN, f1ks), fs) ) return harmonicEuler2(sN, fs, t, res.x[0], f0k1, f0kN, f0ks, res.x[1], f1k1, f1kN, f1ks), res.x[0], res.x[1]#[0] def guessf0( sN, fs ): S = np.fft.fft(sN) w = np.fft.fftfreq( len(sN), 1/fs ) imax = np.argmax( np.abs(S) ) #plt.plot( w, np.abs(S) ) #plt.show() #print(w) #print ( w[imax], w[imax+1] ) return abs(w[imax]) if __name__ == "__main__": import matplotlib.pyplot as plt f0 = 60 # Hz f1 = 60 # Hz delta = np.random.rand() - .5 delta2 = np.random.rand() - .5 print("delta", delta) print("delta2", delta2) fs = 10000 # GMR t = np.arange(0, 1, 1/fs) phi = 2.*np.pi*np.random.rand() - np.pi phi2 = 2.*np.pi*np.random.rand() - np.pi print("phi", phi, phi2) A = 1.0 A2 = 0.0 A3 = 1.0 nK = 10 T2 = .200 sN = A *np.sin( ( 1*(delta +f0))*2*np.pi*t + phi ) + \ A2*np.sin( ( 1*(delta2 +f1))*2*np.pi*t + phi2 ) + \ np.random.normal(0,.1,len(t)) + \ + A3*np.exp( -t/T2 ) sNc = A *np.sin( (1*(delta +f0))*2*np.pi*t + phi ) + \ A2*np.sin( (1*(delta2+f1))*2*np.pi*t + phi2 ) + \ + A3*np.exp( -t/T2 ) guessf0(sN, fs) # single freq #h = harmonicEuler( f0, sN, fs, nK, t) h = minHarmonic( f0, sN, fs, nK, t) # two freqs #h = minHarmonic2( f0+1e-2, f1-1e-2, sN, fs, nK, t) #h = harmonicEuler2( f0, f1, sN, fs, nK, t) plt.figure() plt.plot(t, sN, label="sN") #plt.plot(t, sN-h, label="sN-h") plt.plot(t, h, label='h') plt.title("harmonic") plt.legend() plt.figure() plt.plot(t, sN-sNc, label='true noise') plt.plot(t, h, label='harmonic removal') plt.plot(t, np.exp(-t/T2), label="nmr") plt.legend() plt.title("true noise") plt.show()