Rubidium provides small footprint solutions as embedded software libraries and APIs that can run on many third party commercial core processors in the market.
When implementing a Voice User Interface in any consumer or industrial device, integrating Rubidium's software decreases time to market, saves PCB space, power use and BOM expense, while covering virtually all speech and voice dialog requirements.
The following speech processing technologies are available as software libraries:
- ASR [Automatic Speech Recognition]
- TTS [Text to Speech]
- BSV [Biometric Speaker Verification]
- CSP [Compressed Speech Playback/Storage]
We supply low cost, small footprint solutions by providing code porting, optimization, integration, customization and any other engineering requirements specific to your microprocessor or hardware platform. Because our licensing model is flexible and can be based on units delivered and Non-Recurring Engineering (NRE), your R&D and manufacturing costs are kept to a minimum.
Unlike companies which rely on third party products, all of Rubidium's technology is developed in-house by a team of experts with years of hands-on scientific and industrial experience. The technology is available as chips or software libraries. This provides our clients with flexibility to select the most effective solution, optimized to fit their current and future speech processing needs.
Rubidium's embedded software is deployed in products produced by some of the world's leading brands. Over 25 million products shipped to the marketplace, from Hong Kong to New York, can attest to the quality of our solutions as well as our clients' success in selecting Rubidium as their supplier.
Rubidium's software solutions were successfully ported on CSR's BlueCore5 MM, ARM7, ARM9, TeakLite and other microcontrollers, microprocessors and DSPs.
Main application features
ASR - Automated Speech Recognition easily and safely controls any set of functions through voice commands. For example: call acceptance and rejection, device setup and installation procedure (pairing, calibration, interconnection, etc.), voice dialing, music streaming control and music selection.
- Recognizes any speaker and accent right out of the box
- Multi-lingual - supports any language and even a combination of languages(!)
- Can switch vocabularies in real time to follow the dialog's active state
- Adjustable recognition sensitivity and timeout
- Noise robust - suitable for automotive, office and other noisy environments
- Patent-pending barge-in support allows the user to start speaking while voice announcement is still active
- Keyword Spotting and voice-trigger capability
- Zero latency - results provided immediately when the user stops speaking
TTS - Text To Speech announces any name in any language with a natural and human sounding voice:
- Announces any name or number without programming or recording
- Fluent, natural sounding wideband speech, 16KHz sample rate
- Selection of multiple languages, accents, female or male voices
- Adjustable volume, intonation depth and pitch
- Simple and clear voice prompts supported by CSP technology (see below)
- Cost effective software-only solution
- Low effect on battery recharge time due to low resource consumption
- Large internal dictionary ensures correct pronunciation for optimal accuracy
- Uses concatenative TTS technology based on real human speech
BSI - Biometric Speaker Identification engine offers a small footprint and a low-cost design for affordable and effective biometric identification and authentication on any consumer product or mass market device.
- Double identification layers for improved security: compares both voice characteristics and spoken content
- Users can select their own voice password - no need to memorize digit sequences
- Language independent - supports passwords in any language and accent
- Does not need special sensors - results in a more convenient interface and a lower cost than alternative biometric solutions (like fingerprint or iris ID)
- Supports the needs of the visually impaired
- Favored over traditional ID solutions like ID cards, keypads and fingerprints: friendlier, does not require direct body contact, hygienic, easier to mount, etc.
CSP - Speech Compression Algorithm offers storage and playback of pre-recorded speech prompts that can be used to play user instructions, device status indications, alert messages, remaining battery time, entertaining content, memo recordings and any other information.
- Any requirements of pre-recorded voice prompts can be supported to match any set of languages, announcements and voices (even several languages and voices in a single product!)
- High-quality 16KHz wideband speech
- Resource conscious due to powerful low-bit-rate compression, as low as 800 bytes/sec for wideband speech
- Number of prompts in a single product is virtually unlimited
- A large set of prompts is already available in many languages
- For unique prompts, Rubidium provides recording services to support any customer requests
- Voice 'personalities' and celebrity voices can be tailor-made to specific requirements