Speech Synthesis TRIZ Cases
- Distributed Text-to-Speech Processing Cost Scheduling
A distributed text-to-speech system schedules processing requests across multiple servers to reduce operational costs.
- End-to-End Text-to-Speech Synthesis Using Attention Weights
Encoder maps phonemes to states and decoder generates audio waveforms using attention weights, reducing training complexity.
- Sound Processing System Pronunciation Style Switching
A sound processing system alters time-series data portions to generate new pronunciation styles while maintaining user-specified sound characteristics.
- Expressive Prosody Model Generation via Synthetic Neutral Vectors
A machine learning module trains on parallel neutral prosody vectors generated from expressive recordings to produce natural speech audio.
- Speech Synthesis Phase Correction for Natural Prosody
A speech synthesis method preserves phase information using speaker-dependent phase differences to synchronize pitch periods.
- Hybrid TTS Engine Caching Speech Units for Latency and Storage
A hybrid text-to-speech engine segments voice models to cache speech units locally, enabling immediate synthesis without network dependency.
- Speech Synthesizer Using Pre-Computed Band Signals
Pre-computing band noise and pulse signals eliminates real-time filtering, enabling faster waveform generation while maintaining high-quality tone.
- Text-to-Speech Phoneme Insertion for Spontaneous Rhythm
A text-to-speech system inserts additional phonemes into initial sequences to generate speech with varying rhythms.