CodeIgniter调用Google Speech-to-Text报GPBDecodeException错误求解决

问题代码 #

<?php
defined('BASEPATH') or exit('No direct script access allowed');

use Google\Cloud\Speech\V1\RecognitionAudio;
use Google\Cloud\Speech\V1\RecognitionConfig;
use Google\Cloud\Speech\V1\SpeechClient;
use Google\Cloud\Speech\V1\RecognitionConfig\AudioEncoding;
use Google\Cloud\Speech\V1\StreamingRecognitionConfig;

class SpeechToText extends CI_Controller
{
    public function index()
    {
        $this->config->load('google_speech', TRUE);
        $credentialsPath = $this->config->item('google_speech_credentials_path', 'google_speech');

        // Create Speech Client
        $speech = new SpeechClient([
            'credentials' => $credentialsPath
        ]);

        $recognitionConfig = new RecognitionConfig();
        $recognitionConfig->setEncoding(AudioEncoding::FLAC);
        $recognitionConfig->setSampleRateHertz(44100);
        $recognitionConfig->setLanguageCode('en-US');
        $config = new StreamingRecognitionConfig();
        $config->setConfig($recognitionConfig);

        $url = base_url('uploads/file_example_WAV_1MG.wav');
        $audioResource = fopen($url, 'r');

        $responses = $speech->recognizeAudioStream($config, $audioResource);

        foreach ($responses as $element) {
            // doSomethingWith($element);
        }
    }
}

错误信息 #

Type: Google\Protobuf\Internal\GPBDecodeException
Message: Error occurred during parsing: Fail to push limit.

解决方案 #

1. 修正音频编码与文件格式不匹配问题 #

你配置的编码为FLAC，但实际使用的是WAV格式文件，二者不兼容。WAV文件通常采用LINEAR16编码，需修改配置：

$recognitionConfig->setEncoding(AudioEncoding::LINEAR16);

若坚持使用FLAC编码，需先将WAV文件转换为FLAC格式。

2. 使用本地文件路径替代HTTP URL #

fopen读取HTTP URL可能受服务器allow_url_fopen配置限制，且流式识别更适配本地文件。替换为CodeIgniter本地文件路径：

// 替换原$url和$audioResource代码
$audioPath = APPPATH . '../uploads/file_example_WAV_1MG.wav';
$audioResource = fopen($audioPath, 'r');

根据项目目录结构调整路径，确保文件存在且拥有可读权限。

3. 验证音频采样率与配置一致 #

检查WAV文件的实际采样率是否为44100Hz，若不一致，修改setSampleRateHertz的值为文件实际采样率（如常见的16000Hz）。

4. 用非流式识别替代流式识别（适合本地文件） #

如果仅处理本地音频文件，无需使用流式识别recognizeAudioStream，改用普通recognize方法更稳定：

// 替换原流式识别相关代码
$audio = RecognitionAudio::fromArray([
    'content' => file_get_contents($audioPath)
]);

$response = $speech->recognize($recognitionConfig, $audio);
foreach ($response->getResults() as $result) {
    $alternatives = $result->getAlternatives();
    foreach ($alternatives as $alternative) {
        echo '转录结果: ' . $alternative->getTranscript() . "\n";
    }
}

修正后的完整示例（非流式版本） #

<?php
defined('BASEPATH') or exit('No direct script access allowed');

use Google\Cloud\Speech\V1\RecognitionAudio;
use Google\Cloud\Speech\V1\RecognitionConfig;
use Google\Cloud\Speech\V1\SpeechClient;
use Google\Cloud\Speech\V1\RecognitionConfig\AudioEncoding;

class SpeechToText extends CI_Controller
{
    public function index()
    {
        $this->config->load('google_speech', TRUE);
        $credentialsPath = $this->config->item('google_speech_credentials_path', 'google_speech');

        $speech = new SpeechClient([
            'credentials' => $credentialsPath
        ]);

        $recognitionConfig = new RecognitionConfig();
        $recognitionConfig->setEncoding(AudioEncoding::LINEAR16);
        $recognitionConfig->setSampleRateHertz(44100);
        $recognitionConfig->setLanguageCode('en-US');

        $audioPath = APPPATH . '../uploads/file_example_WAV_1MG.wav';
        $audio = RecognitionAudio::fromArray([
            'content' => file_get_contents($audioPath)
        ]);

        try {
            $response = $speech->recognize($recognitionConfig, $audio);
            foreach ($response->getResults() as $result) {
                foreach ($result->getAlternatives() as $alternative) {
                    echo '转录结果: ' . $alternative->getTranscript();
                }
            }
        } catch (\Exception $e) {
            echo '错误信息: ' . $e->getMessage();
        } finally {
            $speech->close();
        }
    }
}

内容的提问来源于stack exchange，提问作者user24973246