SPEECH RECOGNITION PROJECT

Logical Designs developed a phoneme based speech recognition system and customized preprocessing. Running in real time, the digitized audio signal was preprocessed FFT and customized adaptive gain control technology into inputs to a set of neural network phoneme detectors. Output from the phoneme detectors was then used with a phoneme dictionary to produce the text output.

What makes this methodology interesting is the ability to avoid the most common problem facing speech recognition. This problem is the segmentation vs. recognition problem. Most recognizers must segment the audio stream into words or phonemes prior to recognition. Then the segmented data is further processed and passed to word or phrase recognizers. Logical Designs phoneme detectors run continuously and the output is used in combination with the dictionaries to form the output text. No segmentation eliminates the need for unnatural gaps between words in dictation and ultimately, higher recognition rates.

Development of the Logical Designs Speech techniques was performed using the TIMIT speech database for training set development.

Logical Designs has developed neural algorithms not available from any other source. We can supply a custom software and hardware solution to your problem, or work with the tools you have in house. Logical Designs has the ability and experience to make your project and application a success.