Acoustic target identification with AR spectral estimation based on SVD-TLS

To improve anti-noise performance of AR spectral estimation, an algorithm based on SVD-TLS (singular values decomposition and total least squares) is introduced. With an example the algorithm is compared with other algorithms of AR spectral estimation, such as Least Squares algorithm, Yule-Walker algorithm and Burg algorithm. By using SVD-TLS some harmonic frequencies and amplitudes are extracted from really measured helicopter acoustic signals. All results show that SVD-TLS is superior to other algorithms in extracting frequency information from acoustic signals under the condition of lower SNR and this advantage is very important to the system of acoustic target identification.