A NEW WAY TO ANALYZE SPEECH

Zanavox has developed effective new methods for analyzing speech.   Based on novel correlations in speech wave patterns, the new technology provides extremely fast, reliable command recognition.   Unlike prior speech-recognition technology, the new algorithms use only low-cost microcontrollers and consume negligible power.  

The new technology leads to numerous applications in the markets of voice-activated instrumentation, human-guided automation, speech compression, and (as a preprocessor) natural language interpretation.  

Following is a summary of the Zanavox patent portfolio.   Disclosed are detailed methods and new applications based on Zanavox technology.   All patents are completely unencumbered, fully owned by Zanavox, and are now ready for licensing or sale.   Click each patent number to download a PDF.  

We now seek licensing partners with the ability to implement this technology.   For more information, contact us at Info@Zanavox.com or (760) 522-3463.



  (1.) US8731941: Voice-Activated Measurement System.

The patent covers all voice-activated instruments that perform a measurement responsive to a first command such as "GO", and then transmit or display the measurement result.   The patent further covers a reset function, responsive to "RESET", which returns the display to zero and re-arms the instrument.  

Application examples: electronic test instruments such as oscilloscopes and digital voltmeters, field sensor instruments such as Gaussmeters and electric field sensors, physical parameter sensors such as digital thermometers and lightmeters.   All such instruments can be controlled using the patented method.  


  (2.) US8781821: Voiced Interval Command Interpretation.

Specifies the new Voiced-Interval Method in detail.   Extends patent coverage to any number of intervals.   Also covers toggle-type logic under voice control.  

Advantages over prior art: speaker universality, language independence, easy adaptation of commands to applications, no training or calibration needed, and compatibility with low-cost processors such as embedded microcontrollers.


  (3.) US8862476: Voice-Activated Signal Generator.

This expands the technology to include both voiced and unvoiced intervals in any order.   It also introduces novel signal processing algorithms to accomplish voiced-unvoiced separation with minimal resources.   Output signal may be a pulse, gated wavetrain, level change, toggling function, and other modes.  

Claims provide broad coverage for a range of instruments controlled by voice commands.   Also covered are "enablement" commands, complex logic, channel-selection under voice control, and others.  


  (4.) US8924209: Identifying Spoken Commands by Templates of Ordered Voiced and Unvoiced Sound Intervals.

This patent greatly expands the signal processing protection.   Operational details essential to the fast, low-cost operation are disclosed.   Also covered are effective techniques for noise suppression, particularly noise from voiced-like backgrounds during unvoiced sounds, and vice-versa.   Also covered is the use of a tally criterion for detecting the presence of voiced and unvoiced sound and silence.  

The methods detailed herein also enable fast preprocessing of sounds for speech compression, resulting in faster, more economical transmission.   The claimed methods also contribute to natural-language interpretation by solving the phoneme identification problem upstream of higher-level interpretation routines.  


  (5.) US8954334: Voice-Activated Pulser.

The methods disclosed herein provide enhanced flexibility to voice-controlled instruments, particularly by arranging a "regulator" parameter whose state is adjusted by the speaker.   Operationally, voice commands control a range of responses, depend on the preceding voice commands.  

A pulser is a good example, but is only one of many instrumentation applications enabled by this technology.   For example, the ability to trigger a pre-arranged response, on demand and hands-free, has numerous applications in medicine, especially surgery, and in electronics such as development and repair, plus many other field where a voice-controlled third hand is needed.  


  (6.) US8990079: Automatic Calibration of Command-Detection Thresholds.

This patent discloses a method for an old and unsolved problem: setting a sound detection threshold before the first spoken command is detected.   Separate thresholds for voiced and unvoiced sounds are provided, all completely automatic and in the first few seconds of operation.   The method also provides methods for automatic adjustment of speech sounds based on an agile signal-to-noise tracking protocol.  

Rapid initial threshold setting will enable instruments to recognize speech as soon as they are turned on.   In addition, the methods ensure that a change in speech parameters, or sudden increase in backgrounds, will be quickly recognized and the detection thresholds adjusted accordingly.  


  (7.) US9202463: Voice-Activated Precision Timing.

We have achieved unprecedented sub-millisecond time precision using only spoken commands.   This patent shows how.  

In some versions, the particular response is governed by he type of leading sound - voiced or unvoiced.   This allows the operator to trigger a particular response precisely when needed.   The new sound-processing methods are essential to achieving this time precision.  

Other versions use an "arming" command to sensitize the system, followed by a triggering command.   This results in superb time control for a single response type, or for different response types depending on the partciualr arming command used.  

In addition, a method for disarming an already-armed system is disclosed.   This is important for many applications where a response must be called off at the last possible moment.  

Laboratory tests have demonstrated phenomenal time correlations using the patented methods.   We routinely obtain uncertainties well below 1 millisecond.   The claimed methods are ready to open many applications where a fast, extremely precisely timed response is needed to a spoken command, while also being able to disarm at will.  


  (8.) US9454976: Efficient Discrimination of Voiced and Unvoiced Sounds.

This patent discloses, and protects, a set of extremely effective methods for identifying voiced and unvoiced sounds in speech.   Advanced signal processing is presented, along with experimental proof of its advantages, leading to over 99% accuracy in identifying both voiced and unvoiced spoken sounds among noise.  

In addition, the methods are demonstrated to be fully operational in a primitive 8-bit microcontroller costing mere pennies.  

Efficient voiced-unvoiced discrimination will enable a host of products in communications and instrumentation, and wherever advanced speech sound analysis is needed.  



  CONTACT US

Zanavox Instruments
13446 Poway Road, Suite 236
Poway CA 92064
760-522-3463
Info@Zanavox.com