I. THEORY Doppler radar has been widely applied in the medical domain in the recent years [1]. This non-contact technique is very popular and interesting since no sensor needs to be placed on the body. In a Continuous Wave (CW) Doppler detection system for vital signals, a sinusoidal signal T (t) = A e cos (2πf t) at carrier frequency f is transmitted towards a human body, and then reflected by the chest which moves according to physiological movements. These movements can be detected and evaluated by measuring the frequency shift of the reflected signal according to the well known Doppler-effect law. The reflected signal is demodulated by an IQ quadrature receiver to avoid detection issues due to lack of reception during certain intervals of time [2]. The complex baseband signal is modulated by tiny physiological movements x(t) of the human body. Here, two principal physiological movements (heartbeating and respiration) are represented by a single tone sinusoidal signal, respectively, x(t) = x r (t)+x h (t) = m r sin(ω r t + φ r) + m h sin(ω h t + φ h), where m r and m h describe the amplitudes of respiration and heartbeat movement, respectively, ω r and ω h represent the movement frequencies, φ r and φ h are the initial phases. The exponential term of the reflected baseband signal can be expanded using Fourier series [1], B(t) = +∞ n=−∞ +∞ k=−∞ J n 4πm h λ J k 4πm r λ exp [j (nω h t + kω r t)] exp [j (nφ h + kφ r)] exp (jψ) , (1) where J n is the n-th order Bessel function of the first kind. λ = c/f = 5 mm is the working wavelength of the CW radar at 60 GHz. ψ is defined as the total residual phase of the system. This complex baseband signal in the frequency domain is represented by a sum of harmonic components, which causes not only creation of harmonic frequencies for each physiological movement signal itself, but also generates intermodulated frequencies between the two movements [3]. Due to the nonlinear nature of the Doppler phase modulation, undesired intermodulations of harmonic components are present [1]. A direct peak detection from the spectral analysis is therefore no more reliable, since ambiguities can arise when interpreting the different peak locations. This communication presents an approach to estimate body movements related to vital activities by means of a 60 GHz Doppler radar, using robust optimization algorithms. II. EXPERIMENTAL MEASUREMENT Fig. 1 shows the measured time-domain and frequency-domain breathing signal, superimposed with the heartbeat of a human body, at 1 m distance in front of one emitting and one receiving antenna. The emitting power is about −2.5 dBm and the pyramidal horn antenna gain is 24 dBi. The recording time is 30 seconds. To obtain the spectrum shown in Fig. 1 (b), the acquisition time is taken for 10 seconds, and the sampling frequency is F s = 100 Hz. We can deduce that the first peak in the spectrum in Fig. 1 (b) corresponds to the fundamental of respiration. However, the heartbeat fundamental is not visible, hence the direct spectral analysis does not work at all, as the spectrum depends on several unknown parameters, f r , f h , m r , m h , φ r , and φ h , which persuades us to find a better spectrum-estimation algorithm.