Track-before-detect (TBD) algorithms can improve track accuracy and follow low signal-to-noise ratio targets. A price paid for this increased performance is the high computational complexity of TBD implementations. In this work, we develop a new TBD approach capable of handling raw hydrophone data. In order to learn more about its performance and feasibility when applied to sonar, we use data from the sea trial PreDEMUS'06 with DEMUS sensor array of NATO Undersea Research Centre. As a first step, we introduce the sensor model for a bistatic sonar based on DEMUS receivers. Then, we formulate the TBD problem at hand as a binary hypothesis testing problem and derive a class of adaptive algorithms by using design procedures based upon the generalized likelihood ratio test. Remarkably, such detectors guarantee the constant false track acceptance rate property under the design assumptions with respect to the overall spectral properties of the noise. A preliminary performance analysis is presented. Finally, we discuss its potential to implement automatic track continuation and to prepare automatic classification for temporarily weak targets as these tasks are usually the challenges multistatic sonar systems have to overcome.