SOFTWARE · VOICE · COMMUNICATION

Call-audio diagnosis and speech integration

Separate two purchases: diagnosis and tuning of echo, noise suppression, VAD, AGC, and AEC; and speech-to-text or text-to-speech integration after sample checks. iTalkGo is the reference implementation.

Discuss your project WebRTC integration and call-audio development

Call-audio diagnosis and speech

Diagnose echo, noise suppression, VAD, AGC, and AEC on the call path, or integrate ASR or TTS after a sample check. iTalkGo is the reference product. This is not medical diagnosis.

The reference product is iTalkGo. italkgo.com/en/technology.html

VAD
AGC
AEC
Noise suppression
ASR
TTS
Smart voice assistant
3D surround microphone

Scope

Capture, voice activity detection (VAD), automatic gain control (AGC), acoustic echo cancellation (AEC), neural-network noise reduction, automatic speech recognition (ASR), text-to-speech (TTS), and a smart voice assistant. 3D surround microphone pickup can be scoped for a device. Offline operation is assessed on the target device. iTalkGo is the reference implementation.

Deliverables A working speech pipeline, calling interfaces, deployment notes, and an evaluation on the agreed samples.

For transcription, voice assistants, and interactive terminals

Call-audio diagnosis

Echo, noise, dropouts, and one-way audio are diagnosed on the target product. The report names the cause we can support and the repair scope still to be priced.

Send a fault description, same-condition samples or reproduction steps, and the target device. Do not send passwords or unrelated personal data.

You receive a written diagnosis, the likely cause, and a bounded repair scope. A before/after comparison is used only on the agreed samples.

Audio diagnosis here means engineering diagnosis. It is not a medical diagnosis, and browser demo readings are not published as product scores.

BEFORE WE START

Questions you may want answered before we start.

Does speech recognition run on the device or in the cloud?

That is assessed on the target device. Fully local, cloud, or a mix depends on the processor, memory, language, license, and response time.

Can language-model integration be part of the scope?

Yes. It can sit with speech recognition, text-to-speech, and audio preprocessing. Offline behavior is checked on the target device.

LET’S DISCUSS YOUR PROJECT

Tell us the problem you are solving.

Whether the requirement is already clear, or you want to validate one technical approach first, you can write to us. The target platforms, the current problem, and the result you expect help us judge the development scope.

newsun@tenryutei.com

Japan TENRYU TEI Co., Ltd.
Japan · Naha, Okinawa
tenryutei.com/en/