Discover published sets by community

A statistical model that maps audio signals to phonetic units.

A common metric for evaluating the performance of a speech recognition system, calculated as the number of errors divided by the number of words spoken.

A probabilistic framework that predicts the likelihood of a sequence of words.

An output layer for neural network models that allows the alignment of input audio frames with output labels, often used in end-to-end ASR.

The process of transforming raw audio data into a set of numerical features that can be processed by an ASR system.

An ASR approach that directly maps raw audio to text using a single neural network model, without separate acoustic and language models.

A statistical model used to represent stochastic processes, often employed in ASR to model the probability of transitions between acoustic model states.

The smallest unit of sound in speech that can distinguish one word from another.

The use of computer algorithms to convert spoken language into text.

A search algorithm that expands only a limited set of nodes that are most promising, often used in ASR to decode the spoken words from an acoustic model's output.

A neural network with multiple hidden layers between the input and output layers, used in modern ASR for modeling complex patterns in data.

A technique in ASR to determine the presence or absence of human speech within an audio stream.

A representation of short-term power spectrum of sound, based on a linear cosine transform of a log power spectrum on a nonlinear mel scale of frequency.

The process of partitioning an input audio stream into homogenous segments according to the speaker identity.