604 results about "Timbre" patented technology

A method and apparatus for classifying mood of music at high speed. The method includes: extracting a Modified Discrete Cosine Transformation-based timbre feature from a compressed domain of a music file; extracting a Modified Discrete Cosine Transformation-based tempo feature from the compressed domain of the music file; and classifying the mood of the music file based on the extracted timbre feature and the extracted tempo feature.

The invention provides a voice conversion scheme for non-parallel corpus training to get rid of dependence on parallel texts and solve the technical problem of difficult realization of voice conversion under the condition that resources and equipment are limited. The invention relates to a voice conversion system, method and an application terminal. Compared with the prior art, the scheme has theadvantages that: a trained speaker-independent voice recognition model can be used for any source speaker, namely the speaker is independent; the bottleneck characteristics of the audio are more abstract than voice phonetic posterior Gram characteristics, can reflect the speaking content and decouple from the tone of the speaker, and are not closely bound with the phoneme category, and are not ina clear one-to-one correspondence relationship, so that the problem of inaccurate pronunciation caused by ASR recognition errors is relieved to a certain extent. The pronunciation accuracy of the audio obtained by performing voice conversion by using bottleneck characteristics is obviously higher than that of a voice phonetic posterior Gram method, and the tone is not obviously different; by meansof a transfer learning mode, dependence on training corpora can be greatly reduced.

Multi-channel audio content is mixed for a particular loudspeaker setup. However, a consumer's audio setup is very likely to use a different placement of speakers. The present invention provides a method of rendering multi-channel audio that assures replay of the spatial signal components with equal loudness of the signal. A method for obtaining an energy preserving mixing matrix (G) for mixing L1 input audio channels to L2 output channels comprises steps of obtaining a first mixing matrix Ĝ, performing a singular value decomposition on the first mixing matrix Ĝ to obtain a singularity matrix S, processing the singularity matrix S to obtain a processed singularity matrix Ŝ, determining a scaling factor α, and calculating an improved mixing matrix G according to G=αUŜVT. The perceived sound, loudness, timbre and spatial impression of multi-channel audio replayed on an arbitrary loudspeaker setup practically equals that of the original speaker setup.

The present invention is a method and system to convert speech signal into a parametric representation in terms of timbre vectors, and to recover the speech signal thereof. The speech signal is first segmented into non-overlapping frames using the glottal closure instant information, each frame is converted into an amplitude spectrum using a Fourier analyzer, and then using Laguerre functions to generate a set of coefficients which constitute a timbre vector. A sequence of timbre vectors can be subject to a variety of manipulations. The new timbre vectors are converted back into voice signals by first transforming into amplitude spectra using Laguerre functions, then generating phase spectra from the amplitude spectra using Kramers-Knonig relations. A Fourier transformer converts the amplitude spectra and phase spectra into elementary waveforms, then superposed to become the output voice. The method and system can be used for voice transformation, speech synthesis, and automatic speech recognition.