Interface AudioMetricsDetails

An array of AudioMetricsHistogramBin objects that defines a histogram of the clipping rate for the audio segments. The clipping rate is defined as the fraction of samples in the segment that reach the maximum or minimum value that is offered by the audio quantization range. The service auto-detects either a 16-bit Pulse-Code Modulation(PCM) audio range (-32768 to +32767) or a unit range (-1.0 to +1.0). The clipping rate is between 0.0 and 1.0, with higher values indicating possible degradation of speech recognition.

An array of AudioMetricsHistogramBin objects that defines a histogram of the cumulative direct current (DC) component of the audio signal.

The end time in seconds of the block of audio to which the metrics apply.

If true, indicates the end of the audio stream, meaning that transcription is complete. Currently, the field is always true. The service returns metrics just once per audio stream. The results provide aggregated audio metrics that pertain to the complete audio stream.

The probability that the audio signal is missing the upper half of its frequency content.

• A value close to 1.0 typically indicates artificially up-sampled audio, which negatively impacts the accuracy of the transcription results.
• A value at or near 0.0 indicates that the audio signal is good and has a full spectrum.
• A value around 0.5 means that detection of the frequency content is unreliable or not available.

An array of AudioMetricsHistogramBin objects that defines a histogram of the signal level in segments of the audio that do not contain speech. The signal level is computed as the Root-Mean-Square (RMS) value in a decibel (dB) scale normalized to the range 0.0 (minimum level) to 1.0 (maximum level).

The signal-to-noise ratio (SNR) for the audio signal. The value indicates the ratio of speech to noise in the audio. A valid value lies in the range of 0 to 100 decibels (dB). The service omits the field if it cannot compute the SNR for the audio.

An array of AudioMetricsHistogramBin objects that defines a histogram of the signal level in segments of the audio that contain speech. The signal level is computed as the Root-Mean-Square (RMS) value in a decibel (dB) scale normalized to the range 0.0 (minimum level) to 1.0 (maximum level).

The ratio of speech to non-speech segments in the audio signal. The value lies in the range of 0.0 to 1.0.