THE MAJOR PROBLEM IN FORENSIC SPEAKER RECOGNITION IS THAT STABLE WORKING CONDITIONS CANNOT BE MAINTAINED. SPEAKERS ARE NOT ALWAYS COOPERATIVE SUCH AS IN BOMB THREATS AND HOSTAGE-TAKING OR EXTORTION CASES, TEXT-INDEPENDENT VOICE RECOGNITION METHODS ARE OFTEN REQUIRED, AND VOICE RECORDINGS ARE GENERALLY SUBJECT TO HEAVY INTERFERENCE. A RESEARCH PROJECT IN THE FEDERAL REPUBLIC OF GERMANY IS BEING CONDUCTED TO DEVELOP AN INTERACTIVE FORENSIC SPEAKER RECOGNITION SYSTEM THAT CAN BE PROGRAMMED AND ADAPTED TO VARYING REQUIREMENTS. IN THE MODULAR SYSTEM DESIGN, HARDWARE INCLUDES A COMPUTER AND AN ARRAY PROCESSOR AND SOFTWARE INCLUDES LARGE-SCALE PROGRAM PACKAGES FOR SPEECH SIGNAL ANALYSIS AND PATTERN RECOGNITION. CAPABILITIES OF THE SYSTEM ENCOMPASS SIGNAL PROCESSING, SEGMENTATION, FEATURE EXTRACTION (SHORT-TIME SEGMENT ANALYSIS, CONTOUR ANALYSIS, AND STATISTICAL FEATURE EXTRACTION), STATISTICAL DATA ANALYSIS, AND CLASSIFICATION. EQUATIONS, FIGURES, AND REFERENCES ARE PROVIDED. (DEP)