Corporate / Consumer
Providing Custom Corporate and Consumer Speech Solutions
We Take a Backseat to Your Product
MatrixHCI is in the business of delivering speech solutions that make you look good. That means that our development and our product take a backseat and quietly deliver speech recognition results from within your product line.
For all intents and purposes, your customers will never know we exist.
Corporate Solutions
When standard off-the-shelf speech recognition solutions will not meet your needs, MatrixHCI can develop a speech recognition solution that is custom tailored to perform to your desired specifications.
Speech solutions like Siri, Google, and Nuance have specific functions that they perform and none of them are truly customizable or able to tightly integrate with larger software solutions.
At MatrixHCI, we custom design, tailor, and craft speech solutions that are complete turnkey or are specifically designed to connect and tightly interface with your entire software solution regardless of what programming language you have developed your other applications in.
Also, once developed we can design a plan whereby you own/license your technology with either a per port license fee or with no license fees … whichever works best with your product business model.
We can even host your speech recognition servers in our data center that is specifically designed with speech recognition performance in mind.
Hosted or Stand Alone Self-Contained Speech Services
Depending on your needs, MatrixHCI can develop speech recognition solutions that are Cloud based or solutions that are stand alone residing on a variety of hardware platforms.
Hosted Solutions – We can develop and deploy speech solutions on hosted data servers in our specialized data center or the data center of your choosing.
Stand Alone Solutions – MatrixHCI can also develop stand alone speech solutions on a variety of devices and hardware systems on a wide selection of operating systems.
Our speech solutions have been deployed on smart handheld devices like the iPad Air2, iPad Pro, iPhone 6s series iOS devices, Android devices (with sufficient memory and processing power), Microsoft Surface Pro tablets, and conventional desktop systems running Windows, OS X and Linux.
Whatever your requirements, MatrixHCI can custom design and implement an intelligent speech recognition solution for your specific needs on a platform configuration that best meets your needs.
Custom Acoustic Models and Languages
Because we build our systems from the ground up, we can design a speech recognition solution that specifically meets your needs. Once more, we can design custom acoustic models for any spoken language you need.
We even have the capabilities to collect and record audio samples in specific languages needed, and transcribe the collected audio in preparation to training acoustic models for the spoken languages needed.
Fixed In-House Recording Studios
We have a set of fixed recording studios that are designed to collect both dictational and conversational speech in any language on multiple and different recording sources.
We can collect audio using high quality microphones, conference microphones, iPhone or Android audio, or even audio collected through your own existing or future planned system (i.e., call centers, etc.).
Our recording studios record a cluster of up to 6 different audio input devices simultaneously onto high quality recording devices sampled at 96kHz. This process allows capture of audio in a myriad of different recordings, all with different audio characteristics so that we can best match the audio to your end audio environment. This helps maintain the highest quality of speech recognition for your particular audio acoustical surroundings at your end-product location.
By collecting audio that is close to your actual audio domain, we ensure that the audio collected for training most effectively recognizes with the highest possible accuracy.
Mobile Audio Recording and Data Collection Studios
Sometimes it is necessary to take the audio recording and audio collection studios out into the field in order to obtain very specific audio samples.
MatrixHCI has a converted 40 ft Mobile RV bus that has been retrofitted to function as a stand-alone mobile recording studio or audio collection system.
It is outfitted with a built-in recording studio offering the same multi input capabilities of our in office recording studios and can capture audio of up to 6 different simultaneous input devices at a single time.
This bus can be taken into geographical areas of high concentration of specific languages to capture languages where they are not found in spread out areas of general population.
As an example, it can be taken into a specific geographic area where the desired language population has a high density thus allowing for quicker and easier audio collection in the language desired.
It can also be taken to different geographical parts of the country in order to collect different dialects and accents required for your specific speech recognition needs (i.e., New England, the South, Texas, Mid East States, etc.).
Additionally, the mobile recording studio is also outfitted with 6 custom built antennas designed to collect audio within the general aviation frequencies for collection of aviation and ATC transmissions. Data is captured from these antennas using 6 scanning frequencies.
MatrixHCI has already collected thousands of hours of raw ATC and pilot transmissions from around the country, and we are currently using this data to help build better aviation-related acoustic models.
The mobile recording and audio collection studio is fully self-contained and can sustain long periods of audio recording and data collection using its 7.5 KV on-board diesel generator.
Blended Language Models
We can also collect audio in blended languages like “Spanglish” (Spanish speakers use a combination of English and Spanish words commingled in sentences).
This is important; for example, Spanish primary language speakers who speak English will pronounce Spanish words with Spanish accents and inflection interspersed within normal English pronunciation.
Our experienced team knows how to collect audio and blend it in ways that meet your mixed language speech recognition needs.