Speech/voiceband data discriminator
AssigneeNEC Corporation
InventorsMinoru Shinta, Shinichi Aikoh, Takao Nishitani
Life of this patent
Estimating the term…
Abstract
Input signals are processed to generate a plurality of signals having different features according to whether the input signals are speech signals or voiceband data signals, and these plural signals are entered into a neural network to be determined whether they have features closer to those of speech signals or of voiceband data signals. The classifying function of the neural network is achieved by inputting samples of speech signals and voiceband data signals and learning how to obtain correct classification results.
Claims9
A speech/voiceband data discriminator comprising:
- signal processing means for processing input signals to generate a plurality of signals having different characteristics according to whether the input signals are speech signals or voiceband data signals,
- a neural network responsive to said plurality of signals for generating functional value data to indicate whether the input signals have features close to speech signals or to voiceband data signals, and
- decision means responsive to said functional value data for deciding based on said functional value data whether said input signals are speech signals or voiceband data signals and for supplying signals to indicate the decision, wherein said signal processing means calculates the short time auto-correlation coefficients of said input signals, and generate and send to said neural network a plurality of signals indicating such coefficients,
- wherein said decision means further comprising means for smoothening the data of said functional value with an integrator, and means for comparing the magnitudes of the smoothened data and a predetermined threshold and for outputting a signal indicating whether said input signals are speech signals or voiceband data signals.
A speech/voiceband data discriminator comprising:
- signal processing means for processing input signals to generate a plurality of signals having different characteristics according to whether the input signals are speech signals or voiceband data signals,
- a neural network responsive to said plurality of signals for generating functional value data to indicate whether the input signals have features close to speech signals or to voiceband data signals, and
- decision means responsive to said functional value data for deciding based on said functional value data whether said input signals are speech signals or voiceband data signals and for supplying signals to indicate the decision, wherein said signal processing means calculates the short time auto-correlation coefficients of said input signals, and generate and send to said neural network a plurality of signals indicating such coefficients;
- wherein said decision means further comprising first means for comparing the magnitudes of the data of said functional value and a first predetermined threshold and, integrating means for smoothening the result of said first means for comparing with an integrator, second means for comparing the magnitudes of the smoothened result and a second predetermined threshold and for outputting a signal indicating whether said input signals are speech signals or voiceband data signals.
Dependencies are inferred from the claim text.