Detection device and method for audio direction orientation and audio processing system

An audio processing system, comprising:

• a noise cancellation device, receiving microphone information and speaker play information, and sequentially performing a filtering operation, a sample rate converting operation and an echo cancelling operation on the microphone information and the speaker play information to generate echo-cancelled audio data;
• a beamforming device, coupled to the noise cancellation device to generate a plurality of first audio beams according to the echo-cancelled audio data;
• a direction of arrival (DOA) detection device, receiving the first audio beams, sequentially performing a first infinite impulse response operation, an absolute value operation and a second infinite impulse response operation on the first audio beams to generate a plurality of processed audio beams, and selecting a part of the processed audio beams to generate a plurality of beam information, wherein the beam information correspond to at least one specific direction;
• and
• a speech recognizer, coupled to the DOA detection device, the speech recognizer performs a speech recognition and determines a voice direction according to the beam information.

A detection method for audio direction orientation, comprising:

• receiving a plurality of first audio beams and performing a first infinite impulse response operation on each of the first audio beams to generate a plurality of second audio beams;
• performing an absolute value operation on amplitude of each of the second audio beams to generate a plurality of third audio beams;
• performing a second infinite impulse response operation on each of the third audio beams to smooth each of the third audio beams to generate a plurality of fourth audio beams;
• and
• dividing the fourth audio beams into a plurality of audio beam groups, and selecting a selected audio beam from each of the audio beam groups according to energy of each of the fourth audio beams in each of the audio beam groups so as to output beam information corresponding to the selected audio beams, wherein the beam information are used in a speech recognition and for determining a voice direction.