Fundamentals of speech recognition

By Prof. Homayoon Beigi, H. Beigi. 1993

Subfields: Natural Language Processing, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Speech & Audio

Citations in the corpus

WaveNet: A Generative Model for Raw Audio (2016, background): “However, some phones such as stops are time-limited by less than 20 milliseconds (Rabiner & Juang, 1993).”
Weakly Supervised Action Labeling in Videos under Ordering Constraints (2014, background): “Dynamic time warping algorithms (DTW) can be used to match temporal sequences, and are extensively used in speech recognition, e.g. [7,25].”
Weakly-Supervised Alignment of Video with Text (2015, background): “In speech processing and various related tasks [34], the warping paths are often constrained, forcing for example the path to fall in the Sakoe-Chiba band or in the Itakura parallelogram [40].”
Derivative Dynamic Time Warping (2001, methodology): “Dozens of different step-patterns have been considered, Rabiner and Juang (1993) contains a review.”
On signal reconstruction without phase (2006, methodology): “Most ASR systems use different kinds of cepstral coefﬁcient statistics (see [5,15]) as described next.”
Applications of support vector machines to speech recognition (2004, unknown): “The acoustic modeling components of a speech recognizer are based on hidden Markov models (HMMs) [1], [2].”
Audio-Visual Automatic Speech Recognition: An Overview (2004, methodology): “For LVCSR, a large number of context-dependent sub-phonetic units are used, obtained by clustering the possible phonetic contexts (tri-phone ones, for example) by means of a decision tree (Deller et al., 1993; Rabiner and Juang, 1993; Young et al., 1999).”
An Information-Theoretic Analysis of Hard and Soft Assignment Methods for Clustering (1997, methodology): “These algorithms have been applied in a wide variety of settings, including parameter estimation in hidden Markov models for speech recognition [8], estimation of conditional probability tables in belief networks for probabilistic inference [6], and various clustering problems [3].”
A survey on off-line Cursive Word Recognition (2002, background): “[43,46] and for their speci/c use in handwriting recognition see Ref.”
Statistical methods for speech recognition (1997, background): “Readers already familiar with Rabiner and Juang (1993) will find it an excellent companion volume.”
Extreme deconvolution: Inferring complete distribution functions from noisy, heterogeneous and incomplete observations (2009, background): “…of finding clusters and/or overdensities in the distribution is a problem of significant general interest in many areas of science and of astronomy in particular (e.g., McLachlan and Basford 1988; Rabiner and Biing-Hwang 1993; Dehnen 1998; Helmi et al. 1999; Skuljan et al. 1999; Hogg et al. 2005).”
Recognizing Complex Events Using Large Margin Joint Low-Level Event Model (2012, methodology): “Thus, for solving the video classification we propose to learn low-level events along with their correlations by analyzing the video sequence temporally and using a set of diverse features: ISA (independent subspace analysis) [5], STIP (spatio-temporal interest point descriptor) [19], Dollar [20], G”
On Prediction Using Variable Order Markov Models (2004, background): “Examples of such applications are biological sequence analysis (Bejerano & Yona, 2001), speech and language modeling (Schütze & Singer, 1994; Rabiner & Juang, 1993), text analysis and extraction (McCallum et al., 2000) music generation and classification (Pachet, 2002; Dubnov et al., 2003),…”
New insights into the noise reduction Wiener filter (2006, methodology): “9, where besides and , we also plotted the Itakura–Saito (IS) distance, a widely used objective quality measure that performs a comparison of spectral envelopes (AR parameters) between the clean and the processed speech [53].”
Continuous myoelectric control for powered prostheses using hidden Markov models (2005, methodology): “The iterative expectation-maximization (EM) algorithm [3] was not”
Phase Retrieval By Projections (2013, background): “Examples of this problem occur in speech recognition [7, 20, 22], quantum state tomography [21], and optics applications such as X-rays, crystallography, and electron microscopy [5, 17, 18].”
Survey on speech emotion recognition: Features, classification schemes, and databases (2011, background): “Since human perception of pitch does not follow a linear scale [103], the filters’ bandwidths are usually evenly distributed with respect to a suitable nonlinear frequency scale such as the Bark scale [103], the Mel-frequency scale [103,61], the modified Mel-frequency scale, and the ExpoLog scale [1”
Video Event Understanding Using Natural Language Descriptions (2013, methodology): “3 L(a, r; w) = n i=1 log p(a i , r i ; w) (3) Features: The global video feature uses multiple channels through HOG3D [10], ASR, OCR, MFCC [20] and SIFT [15] features.”
Dynamic Bayesian Networks for Audio-Visual Speech Recognition (2002, methodology): “The parameters of the HMM [6] are π(i) = P q 1 = i , b t (i) = P O t |q t = i , a(i| j) = P q t = i|q t−1 = j , (1) where q t is the state of the backbone node at time t, π(i) is the initial state distribution for state i, a(i| j) is the state transition probability from state j to state i, and b t…”