Paper data
Title:
SSBCarrier Mismatch Detection from Speech Characteristics: Extension beyond the Range of Uniqueness Author(s): Heute Ulrich, LNS/TF/Univ. of Kiel Guelzow Thomas, LNS/TF/ Univ. of Kiel,D24143 Kiel, Germany Kolb HansJoachim, MEDAV GmbH,D91080 Uttenreuth,Germany Page numbers in the proceedings: Volume I pp 495498 Session: Echo Cancellation and Speech Enhancement
Paper abstract
In certain radiosurveillance applications, speech signals with uncertain carrier frequencies may be detected. Beyond other disturbances, carrier mis matches distort the demodulated SSB signals by frequency shifts Df.
Recently, a detection and correction of such errors was presented [1], exploit ing the harmonic structure of voiced speech with a fundamental (or "pitch") frequency fp. Noiserobust, modified pitchdetection methods [2,3] were shown to be applicable, together with an iteratively corrected "coherent superposition" algorithm [4,5]. Thereby, carrier errors are corrected reliably [1]  though only within an apriori uniqueness range Df < fp/2.
In the paper, the extension beyond this limitation is presented. It emerges from a closer, statistical analysis of the measurement errors: The measurement statistics within this range are interpreted as a conditional probability densi ty function (pdf) pDf (xy) conditioned on the actual pitch frequency, its combination with the (measurable) pitchfrequency pdf pfp(y) yields the joint pdf pDf,fp (x,y), and its integration gives the marginal pdf pDf (x). These considerations lead to an even narrower reliability range Df < fp/3, first. The interpretation of the uniqueness range as one period of an fpperiodic repetition of the measurement statistics, however, leads to the solution: After the same steps as sketched above, the marginal pdf pDf (x) shows a strong peak at the true value Df without a (theoretical) restriction. The resulting algor ithm was successfully tested with simulated and real data.
Paper
