Fast waveform synchronization for concentration and time-scale modification of speech
AssigneeNuance Communications, Inc.
InventorsGeert Coorman, Bert Van Coile
Life of this patent
Estimating the term…
Abstract
A synthesis method for concatenative speech synthesis is provided for efficiently concatenating waveform segments in the time-domain. A digital waveform provider produces an input sequence of digital waveform segments. A waveform concatenator concatenates the input segments by using waveform blending within a concatenation zone to synchronize, weight, and overlap-add selected portions of the input segments to produce a single digital waveform. The synchronizing includes determining a minimum weighted energy anchor in the selected portion of each input segment and aligning synchronization peaks in a local vicinity of each anchor.
Claims50
A digital waveform concatenation system for use in an acoustic processing application, the system comprising:
- a digital waveform provider that produces an input sequence of at least two digital waveform segments, each waveform segment being a sequence of samples;
- and
- a waveform concatenator that:
- i. synchronizes input waveform segments to form a sequence of partially overlapping waveform segments, and
- ii. weights and adds selected portions of the overlapping waveform segments to concatenate the input waveform segments so as to produce a single digital waveform;
- wherein for segments of voiced speech, the synchronizing includes aligning a minimum energy anchor in each waveform segment with a corresponding minimum energy anchor of an adjacent waveform segment, each minimum energy anchor location in a given segment being optimized based on determining minimum weighted energy in a neighborhood of a boundary of the given segment.
A digital waveform concatenation system for use in an acoustic processing application, the system comprising:
- a digital waveform provider that produces an input sequence of at least two digital waveform segments, each waveform segment being a sequence of samples;
- and
- a waveform concatenator that:
- i. synchronizes successive waveform segments to form a sequence of partially overlapping waveform segments, the overlapping portion of each waveform segment including an optimization zone near a waveform segment boundary, and
- ii. weights, and adds selected portions of the input segments to concatenate the input segments so as to produce a single digital waveform;
- wherein for segments of voiced speech, the synchronizing includes aligning a largest waveform peak or trough in the optimization zone of each input waveform segment with a corresponding largest waveform peak or trough in an optimization zone of an adjacent waveform segment.
A digital waveform concatenation system for use in an acoustic processing application, the system comprising:
- a digital waveform provider that produces an input sequence of at least two digital waveform segments, each waveform segment being a sequence of samples;
- and
- a waveform concatenator that:
- i. synchronizes successive waveform segments to form a sequence of partially overlapping waveform segments, and
- ii. weights and adds selected portions of the overlapping waveform segments to concatenate the input waveform segments so as to produce a single digital waveform;
- wherein for segments of voiced speech, the synchronizing includes aligning synchronization peaks or troughs in selected portion of each input waveform segment with synchronization peaks or troughs in a corresponding selected portion of an adjacent waveform segment, the location of the selected portions being determined by searching in a neighborhood of waveform segment boundaries for a location where the sum of the weighted energy of the selected portions is minimal.
A digital waveform concatenation system for use in an acoustic processing application, the system comprising:
- a digital waveform provider that produces an input sequence of at least two digital waveform segments, each waveform segment being a sequence of samples;
- and
- a waveform concatenator that:
- i. synchronizes successive waveform segments to form a sequence of partially overlapping waveform segments, and
- ii. weights, and adds selected portions of the overlapping waveform segments to concatenate the input waveform segments so as to produce a single digital waveform;
- wherein for pairs of overlapping segments of voiced speech, a first selected portion includes a minimum energy anchor in a location optimized based on determining minimum weighted energy in a neighborhood of the waveform segment boundaries, and a second selected portion is determined by aligning synchronization peaks or troughs in the neighborhood of the waveform segment boundaries.
Dependencies are inferred from the claim text.