JP2019113636A - Voice recognition system - Google Patents

Abstract

To detect a hot word even in the middle of voice recognition processing after the hot word is detected.SOLUTION: A voice recognition system 1 comprises: a recording buffer from which recorded voice data is outputted; and a primary buffer where the voice data of the recording buffer is copied. The hot word is detected by using the voice data of the primary buffer. Natural language understanding is performed by using the voice data of the recording buffer. Hot word detection and natural language understanding are simultaneously performed after the hot word is detected.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a speech recognition system that performs speech recognition.

  A voice recognition system that performs voice recognition includes, for example, an electronic device such as a speaker device and a cloud server. The electronic device comprises a microphone and a speaker. The speech recognition system recognizes speech input from a microphone, executes processing based on the recognized speech, and outputs an execution result from a speaker. For example, when the user utters "Teach me the weather", an electronic device included in the speech recognition system outputs a voice "The weather today is fine." Patent Document 1 discloses an invention of changing a communication rate as a result of speech recognition.

  In a speech recognition system that performs speech recognition, when a hotword (eg, "Hello, Onkyo") for enabling speech recognition is detected, subsequent speech recognition processing (natural language understanding: NLU (Natural) language understanding)). FIG. 3A is a diagram for explaining hot word detection. The recorded voice data is output to the recording buffer. Hotword detection is performed using audio data of the recording buffer. FIG. 3B is a diagram for explaining processing after hot word detection. After hot word detection, natural language understanding is performed using voice data of the recording buffer.

US Published Patent 2008/0300025

  In the conventional electronic device, there is a problem that hot word detection can not be performed until natural language understanding is finished after hot word detection. For example, it is assumed that the user utters "Tell the weather" after uttering the hot word "Hello, Onkyo". At this time, if the user wants to listen to the music instead of listening to the weather information, "Hello, Onkyo", "Tell the weather", then immediately "Hello, Onkyo", "Music I can not listen to music even if I say "I take it".

  An object of the present invention is to enable hot word detection even during speech recognition processing after hot word detection.

  A voice recognition system according to a first aspect of the present invention comprises a recording buffer to which recorded voice data is output, and a primary buffer to which voice data of the recording buffer is copied, and using the voice data of the primary buffer, Hot word detection is performed, and natural language understanding is performed using speech data of the recording buffer.

  In the present invention, hot word detection is performed using the audio data of the primary buffer to which the audio data of the recording buffer has been copied. In addition, natural language understanding is performed using voice data of a recording buffer. Therefore, since hot word detection and natural language understanding can be performed simultaneously, hot word detection can be performed even during natural language understanding processing after hot word detection.

  A speech recognition system according to a second aspect of the present invention is characterized in that, in the speech recognition system according to the first aspect, hot word detection and natural language understanding are simultaneously performed after the hot word is detected.

  A speech recognition system according to a third aspect of the invention is characterized in that in the speech recognition system according to the first or second aspect, hot word detection is performed using speech data of the recording buffer before a hot word is detected. I assume.

  A speech recognition system according to a fourth aspect of the present invention is the speech recognition system according to any of the first through third aspects, wherein after the hot word is detected, the audio data of the recording buffer is copied to the primary buffer. It features.

  According to the present invention, hotword detection can be performed even during natural language understanding processing after hotword detection.

It is a figure showing the composition of the speech recognition system concerning the embodiment of the present invention. It is a figure showing the composition of the speech recognition system concerning the embodiment of the present invention. (A) It is a figure for demonstrating a hot word detection. (B) is a figure for demonstrating the process after a hot word detection.

  Hereinafter, embodiments of the present invention will be described. FIG. 1 is a block diagram showing the configuration of a speech recognition system according to an embodiment of the present invention. The speech recognition system 1 includes an electronic device and a cloud server. Although not shown, the electronic device includes an SoC (System on Chip), a microphone, a speaker, and the like. The SoC (control unit) controls each unit constituting the electronic device. In the present embodiment, the voice recognition system 1 is configured by an electronic device and a cloud server, and the electronic device and the cloud server cooperate to perform voice recognition.

  The microphone picks up the sound. The sound collected by the microphone is recorded (voice recording). The audio data recorded by the SoC is output to the recording buffer. The detection of the hotword (trigger word) is performed using the audio data of the recording buffer (see FIG. 3A). The hotword is, for example, "Hello, Onkyo". In this embodiment, the SoC performs hotword detection, but hotword detection may be performed by a cloud server. After detection of the hotword, as shown in FIG. 1, audio data of the recording buffer is copied to the primary buffer by the SoC. The audio data of the primary buffer is subsequently used for hot word detection. At the same time, natural language understanding is performed using speech data of the recording buffer. In the present embodiment, natural language understanding is performed on the cloud server.

  In the present embodiment, the electronic device is provided with the recording buffer and the primary buffer, but may be provided in the cloud server.

  As described above, since hot word detection and natural language understanding are simultaneously performed, hot word detection can be performed even during natural language understanding processing after hot word detection. That is, if you immediately say "hello, on kiyo", "put on music" immediately after "hello, on kiyo" or "tell the weather", then hotword detection even while natural language understanding processing is in progress. You can play the music.

  After natural language understanding, speech synthesis is performed. For example, it is assumed that the user utters "Hello, Onkyo", "Tell me about the weather". Corresponding to this, for example, "Today's weather is fine" is output from the speaker. Here, during the speech synthesis processing, as shown in FIG. 2, hot word detection is performed using the speech data of the recording buffer. "Today's weather is fine." During speech synthesis, if "Hello, Onkyo" is uttered, hotword detection is performed using the audio data of the recording buffer.

  As described above, in the present embodiment, hot word detection is performed using the audio data of the primary buffer to which the audio data of the recording buffer has been copied. In addition, natural language understanding is performed using voice data of a recording buffer. Therefore, since hot word detection and natural language understanding can be performed simultaneously, hot word detection can be performed even during natural language understanding processing after hot word detection.

  As mentioned above, although embodiment of this invention was described, the form which can apply this invention is not restricted to the above-mentioned embodiment, It is possible to add a change suitably in the range which does not deviate from the meaning of this invention is there.

  The present invention can be suitably employed in a speech recognition system that performs speech recognition.

1 Speech recognition system

Claims (4)

A recording buffer for outputting the recorded voice data;
A primary buffer to which audio data of the recording buffer is copied;
Equipped with
Hotword detection is performed using voice data of the primary buffer,
A speech recognition system characterized in that natural language understanding is performed using speech data of the recording buffer.   The speech recognition system according to claim 1, characterized in that hot word detection and natural language understanding are performed simultaneously after the hot word is detected.   The speech recognition system according to claim 1 or 2, wherein hot word detection is performed using voice data of the recording buffer before a hot word is detected.   A speech recognition system according to any of the preceding claims, wherein speech data of the recording buffer is copied to the primary buffer after a hotword is detected.

Images