# US 2026/0088015 A1 · Speech synthesis device, speech synthesis method, and computer program product

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0088015 A1
- Document ID: US20260088015A1
- Status: Published application
- Kind: A1
- Application number: 19/407,471
- Filed: 2025-12-03
- Published: 2026-03-26
- Earliest claimed: 2023-09-19
- Assignees: Kabushiki Kaisha Toshiba; Toshiba Digital Solutions Corporation
- Applicants: Kabushiki Kaisha Toshiba; Toshiba Digital Solutions Corporation
- Inventors: Yoshiki HIRUTA; Masatsune TAMURA
- Web page: https://patentgrep.com/patent/US20260088015A1

## CPC

- G10L 13/047: Architecture of speech synthesisers
- G10L 13/10: Prosody rules derived from text; Stress or intonation
- G10L 13/02: Methods for producing synthetic speech; Speech synthesisers
- G10L 25/30: using neural networks
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

A speech synthesis device according to an embodiment includes a memory and a hardware processor connected to the memory. The processor executes encoder processing with a first neural network to convert attribute information of a speech unit into an intermediate representation. The processor executes decoder processing with a second neural network to generate an acoustic feature from the intermediate representation. The processor executes adjustment processing by using an adjustment dictionary in which at least the attribute information of the speech unit is set as a key and an adjustment instruction to the acoustic feature is set as a value. The processor executes the adjustment processing by defining, by the key, a section to which the adjustment instruction is applied, and adjusting the acoustic feature in the defined section based on the adjustment instruction.

## Claims

### Claim 1

A speech synthesis device comprising:
a memory; and
a hardware processor connected to the memory and configured to
execute encoder processing with a first neural network to convert attribute information of a speech unit into an intermediate representation,
execute decoder processing with a second neural network to generate an acoustic feature from the intermediate representation, and
execute adjustment processing by
using an adjustment dictionary in which at least the attribute information of the speech unit is set as a key and an adjustment instruction to the acoustic feature is set as a value,
defining, by the key, a section to which the adjustment instruction is applied, and
adjusting the acoustic feature in the defined section based on the adjustment instruction.

### Claim 2 (depends on 1)

The speech synthesis device according to claim 1, wherein the key of the adjustment dictionary includes information identifying the intermediate representation obtained by the first neural network.

### Claim 3 (depends on 2)

The speech synthesis device according to claim 2, wherein
the intermediate representation is a latent representation obtained by the encoder processing, and
the information identifying the intermediate representation is an index obtained by using a machine learning model that classifies the intermediate representation or the attribute information of the speech unit.

### Claim 4 (depends on 3)

The speech synthesis device according to claim 3, wherein
the machine learning model that classifies the intermediate representation is a clustering model, and
the index is a cluster number obtained when the intermediate representation is classified by the clustering model.

### Claim 5 (depends on 3)

The speech synthesis device according to claim 3, wherein
the machine learning model that classifies the attribute information of the speech unit is a decision tree model, and
the index is a leaf node number reached when the attribute information of the speech unit is input to the decision tree model.

### Claim 6 (depends on 1)

The speech synthesis device according to claim 1, wherein the adjustment instruction is an instruction to replace a vector indicating the intermediate representation with a specified vector.

### Claim 7 (depends on 6)

The speech synthesis device according to claim 6, wherein the instruction to replace with the specified vector is defined for each type of the acoustic feature.

### Claim 8 (depends on 7)

The speech synthesis device according to claim 7, wherein the type of the acoustic feature includes at least one of duration of the speech unit, a logarithmic fundamental frequency, energy, or a spectral feature.

### Claim 9 (depends on 1)

The speech synthesis device according to claim 1, wherein
the acoustic feature is a logarithmic fundamental frequency, and
the adjustment instruction is an instruction to apply a specified operation to the logarithmic fundamental frequency.

### Claim 10 (depends on 1)

The speech synthesis device according to claim 1, wherein
the acoustic feature is duration of the speech unit, and
the adjustment instruction is an instruction to apply a specified operation to the duration.

### Claim 11 (depends on 1)

The speech synthesis device according to claim 1, wherein
the acoustic feature is energy, and
the adjustment instruction is an instruction to apply a specified operation to the energy.

### Claim 12 (depends on 1)

The speech synthesis device according to claim 1, wherein
the acoustic feature is a spectral feature, and
the adjustment instruction is an instruction to apply a specified operation to the spectral feature.

### Claim 13 (depends on 1)

The speech synthesis device according to claim 1, wherein
the acoustic feature is an aperiodic index, and
the adjustment instruction is an instruction to apply a specified operation to the aperiodic index.

### Claim 14

A speech synthesis method implemented by a computer, the method comprising:
executing encoder processing with a first neural network to convert attribute information of a speech unit into an intermediate representation;
executing decoder processing with a second neural network to generate an acoustic feature from the intermediate representation; and
executing adjustment processing by
using an adjustment dictionary in which at least the attribute information of the speech unit is set as a key and an adjustment instruction to the acoustic feature is set as a value,
defining, by the key, a section to which the adjustment instruction is applied, and
adjusting the acoustic feature in the defined section based on the adjustment instruction.

### Claim 15

A computer program product comprising a non-transitory computer readable recording medium on which programmed instructions executable by a computer are recorded, the instructions causing the computer to perform processing, the processing including:
executing encoder processing with a first neural network to convert attribute information of a speech unit into an intermediate representation;
executing decoder processing with a second neural network to generate an acoustic feature from the intermediate representation; and
executing adjustment processing by
using an adjustment dictionary in which at least the attribute information of the speech unit is set as a key and an adjustment instruction to the acoustic feature is set as a value,
defining, by the key, a section to which the adjustment instruction is applied, and
adjusting the acoustic feature in the defined section based on the adjustment instruction.
