• Time alignment and similarity measures became the central paradigm for speech recognition, with Dynamic Time Warping (Dynamic Time Warping, DTW) enabling time-normalized matching and dynamic programming-based time-warping guiding word-level alignment across utterances. [9], [16], [7], [12], [10].

• Front-end representations grounded in linear prediction, cepstral analysis, and formant/pitch estimation provided compact, trainable features for recognition, enabling predictive coding and excitation modeling via Linear Predictive Coding (LPC) and cepstrum-based methods. [1], [19], [4], [5], [3].

• Vector quantization and statistical pattern recognition shaped early automatic speech recognition pipelines, with Vector Quantization (VQ) design, LPC-front ends, and emerging integration of Hidden Markov Models for speaker-independent isolated word recognition. [6], [17], [20], [8].

• Acoustic-phonetic cue research established spectral formants, cepstral pitch, and voicing cues as foundational signals for speech perception and recognition, informing feature extraction and decision rules. [5], [4], [15], [2].