摘要:
A system for enhancing low frequency spectral content of a signal transmitted via a channel, that includes a noise suppression circuit to update channel energy estimates; and a spectral enhancer circuit coupled to and follows the noise suppression circuit, the spectral enhancer circuit to determine channel enhancement in response to the channel energy estimates.
摘要:
The speech recognition training unit is modified to store digitized speech samples into a speech database that can be accessed at recognition time. The improved recognition unit comprises a noise analysis, modeling, and synthesis unit which continually analyzes the noise characteristics present in the audio environment and produces an estimated noise signal with similar characteristics. The recognition unit then constructs a noise-compensated template database by adding the estimated noise signal to each of the speech samples in the speech database and performing parameter determination on the resulting sums. This procedure accounts for the presence of noise in the recognition phase by retraining all the templates using an estimated noise signal with similar characteristics as the actual noise signal that corrupted the word to be recognized. This method improves the likelihood of a good template match, which increases the recognition accuracy.
摘要:
A method and apparatus for canceling both earseal and hybrid echo in a Wireless Local Loop telephone system is disclosed. The present invention operates within a subscriber station having a plurality of telephone inputs, one of which is a dedicated handset and at least one other input being a standard analog telephone. During a conference call between a far end speaker and two near end speakers, the far end speaker may hear an echo of his own voice due to an earseal echo generated by the dedicated handset and a hybrid echo generated by a 4-to-2 wire hybrid interface within the subscriber station. In accordance with the present invention, two distinct echo cancellers are used in the subscriber station to reduce the two types of echo. A first echo canceller is optimized to remove the earseal echo generated at the dedicated handset while a second echo canceller is optimized to remove the hybrid echo generated at the 4-to-2 wire hybrid. The two echo cancellers operate independently such that neither echo canceller need adapt its filter coefficients when a change in the echo channel characteristics occur, such as when a speaker hangs up or another speaker joins the conversation.
摘要:
An improved method and apparatus for canceling echo is disclosed. The echo canceller of the present invention compensates for the existence of quantization error in the echo signal that does not exist in the reference signal by introducing the quantization effects into the reference signal. The quantization effects are typically companding quantization errors caused by the A-law and .mu.-law quantizers prevalent in digital communications equipment.