Change search
ReferencesLink to record
Permanent link

Direct link
Signal processing applied to acoustic distance estimation and sound classification
Luleå tekniska universitet.
1993 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

The thesis treats some signal processing applications to retrieval of information from acoustic measurements. In the distance estimation applications (parts AI, A2, B and C) an acoustic signal propagates through a physical system and is modified according to some system parameter. In the classification problem (parts D1 and D2) the acoustic signal is modified due to the circumstances of its generation. Finally, the design of a signal processor (part E) is described. In the pulse-echo method the parameter to estimate is the propagation time of a reflected acoustic signal. Due to the narrowband bandpass characteristic of ultrasound transducers, the echoes are represented by relatively long duration responses. When multiple echoes from surfaces closely spaced in range are obtained, they are partly overlapping, which causes a resolution problem. A general solution to this problem is to design an excitation signal and a corresponding receiver filter, based on the transducer unit pulse response, to give the shortest possible response at the receiver filter output and simultaneously a minimum acceptable output signal-to-noise ratio. In parts AI and A2 a suboptimal solution has been developed using a combination of resolution enhancement filtering, split between the transmitter and the receiver, and a signal-to-noise ratio increasing pulse compression technique. In part B, a system using airborne focused ultrasound to record a topographic map of a skin surface, is studied. A problem encountered in estimating the time delay for the single echoes from an arbitrarily shaped surface is that the echo shape depends on the local properties of the surface. This introduces a systematic error in the estimate. Dense scanning of the surface and applying a two-dimensional averaging filter to the matrix of range data have reduced the errors to an acceptable level. In part C a method, based on modelling an ore pass as a one-dimensional acoustic tube, has been developed to estimate the position of the border between the empty and the filled parts. The wave propagation in the acoustic tube is modelled as a lattice structure. The parameters of this structure have a direct physical interpretation as reflection coefficients. The border between the empty and the filled parts will correspond to a large reflection coefficient in the lattice model and its location will indicate the ore level. Frequency independent attenuation is incorporated in the model and it is shown how the estimated parameters are affected. A method to separately estimate the forward propagating wave, leads to acceptable results when the model parameters are estimated by an autoregressive model algorithm. In parts D1 and D2 classification methods are applied to secondary sounds emitted by shock wave lithotripsy, used to disintegrate kidney stones inside the body. The objective of the sound classification is to get knowledge of some parameter of the sound generation. In shock wave lithotripsy the acoustic excitation is made outside the body and the waves are focused at the stone location. The positioning of the acoustic focus relative to the stone is critical to achieve the maximum fragmenting effect on the stone. It has been found that the timbre of the audible part of the emitted secondary sound is affected by this. The classification methods have been based on the spectral properties of the sounds. In a method presented in part D1 the parameters of an autoregressive signal model are used as features for classification. A second method developed in part D2 is based on the expansion of logarithmic spectra of the sounds in a set of orthogonal base functions. The coefficients of this expansion are used as features. The base functions are obtained by singular value decomposition of a set of representative spectra. Both methods successfully discriminates between the desired cases, provided that a calibration has been carried out for each new patient. In part E the design of a signal processor, used in conjunction with the work presented in part C, is described. It was designed to accelerate part of the calculations in the algorithm. This design was carried out before standard signal processors became available. Programming tools were also developed.

Place, publisher, year, edition, pages
Luleå: Luleå tekniska universitet, 1993. , 13 p.
Doctoral thesis / Luleå University of Technologyy… → 31 dec 1996, ISSN 0348-8373 ; 115
Research subject
Signal Processing
URN: urn:nbn:se:ltu:diva-26330Local ID: dc0ba590-f67c-11db-ac79-000ea68e967bOAI: diva2:999492

Godkänd; 1993; 20070429 (ysko)

Available from: 2016-09-30 Created: 2016-09-30 Last updated: 2016-10-20Bibliographically approved

Open Access in DiVA

No full text

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

ReferencesLink to record
Permanent link

Direct link