3 results about "Unison" patented technology

This invention belongs to the field of music data processing and relates to a symbolic music classification method and system based on Tonnetz spatiotemporal graphs. The method generates a two-dimensional grid of symbolic music data, consisting of time frames and pitches, and defines graph nodes and initial features for each grid point. Discrete note events are organized into structured nodes. Based on this two-dimensional grid, graph nodes corresponding to pitches that satisfy Tonnetz interval relationships are connected within the same time frame to form spatial edges. These spatial edges explicitly introduce consonant interval relationships between pitches, allowing the model to directly obtain structural information with musicological significance without relying on large amounts of data for self-learning, thus solving the problem of insufficient utilization of harmonic structure. Furthermore, graph nodes with the same pitch or pitch differences within a preset range are connected between adjacent time frames to form temporal edges. These temporal edges directly encode the continuous and stepwise motion of unison parts during voice progression, enabling the model to capture the dynamic evolution of music in the time dimension and solving the problem of missing voice progression characterization.

The application relates to a dynamic error correction method and system for a speech recognition result, which comprises the following steps: performing frame processing on the speech to be recognized, generating a set of phoneme state posterior probability distributions, and calculating a frame-level confusion entropy integral value to quantify local uncertainty; screening low-entropy effective decoding paths based on the frame-level confusion entropy integral value and a signal-to-noise ratio inverse threshold, and then constructing a homophone candidate word lattice network; extracting audio segments corresponding to each node from the network, calculating a fundamental frequency track and a syllable duration, and generating a real-time rhythm feature parameter vector; calling a standard vocabulary pronunciation statistical library to obtain a standard rhythm statistical template, calculating a rhythm feature alignment difference value between the real-time rhythm feature parameter vector and the template, weighting and deducting the confidence of the candidate word according to the rhythm feature alignment difference value, and selecting the vocabulary with the highest confidence as the text after error correction. The application effectively improves the accuracy of speech recognition in a complex acoustic scene by screening decoding paths and introducing rhythm feature information.

The invention relates to a dynamic error correction method and system for a speech recognition result, and the method comprises the steps: carrying out the framing processing of a to-be-recognized speech, generating a phoneme state posterior probability distribution set, and calculating a frame-level confusion entropy integral value to quantify the local uncertainty; screening a low-entropy effective decoding path based on a frame-level confusion entropy integral value and a signal-to-noise ratio inverse ratio threshold value, and further constructing a homophone candidate word grid network; extracting an audio clip corresponding to each node from the network, calculating a fundamental frequency track and syllable duration, and generating a real-time rhythm characteristic parameter vector; and calling a standard vocabulary pronunciation statistical library to obtain a standard rhythm statistical template, calculating a rhythm feature alignment difference value between a real-time rhythm feature parameter vector and the template, carrying out weighted deduction on candidate word confidence, and selecting a vocabulary with the highest confidence as an error-corrected text. According to the invention, the decoding path is screened and the rhythm feature information is introduced, so that the accuracy of speech recognition in a complex acoustic scene is effectively improved.